WEBVTT
00:00:01.919 --> 00:00:14.160
Chapter four Making It Secure The Importance of Cybersecurity Innovation must be balanced by trust to ensure that the intelligence we build cannot be turned against us.
00:00:14.400 --> 00:00:19.679
In this chapter, we explore how cybersecurity must evolve alongside AI.
00:00:19.920 --> 00:00:26.320
As intelligence systems reshape how enterprises operate, new risks demand equally advanced defenses.
00:00:26.480 --> 00:00:34.399
We'll examine emerging threats and the strategies required to secure data, models, and AI-driven operations against them.
00:00:34.719 --> 00:00:50.799
62% of organizations experienced a deep fake attack involving social engineering or exploiting automated processes, while 32% said they experienced an attack on AI applications that leveraged the application prompt in the last 12 months.
00:00:51.119 --> 00:00:59.280
In recent years, cyber threats have evolved from simple breaches to sophisticated assaults aimed directly at enterprise AI systems.
00:00:59.439 --> 00:01:01.679
And the stakes have never been higher.
00:01:01.920 --> 00:01:15.120
As organizations accelerate adoption of technologies such as generative AI, Gen AI, we are witnessing a surge in attacks that leverage AI for phishing, deepfakes, and advanced social engineering.
00:01:15.359 --> 00:01:30.000
At the same time, a new wave of vulnerability is emerging: malicious actors who exploit Gen AI infrastructure, manipulate prompts, or compromise chained AI workflows to infiltrate and disrupt enterprises.
00:01:30.319 --> 00:01:38.799
In the previous chapter, we examined how the intersection of data and enterprise AI creates opportunities for innovation and operational efficiency.
00:01:39.040 --> 00:01:42.079
But opportunities also bring unwanted risks.
00:01:42.239 --> 00:01:51.359
And as organizations rely on private data to fuel their AI engines, they may inadvertently expose themselves to new and evolving cyber risks.
00:01:51.599 --> 00:02:00.480
The protection of enterprise data and AI must keep pace with technological evolution, as data and AI are attractive targets for threat actors.
00:02:00.719 --> 00:02:14.960
In the following feature, an energy company lays the foundation for AI and advanced analytics within a secure EIM system, building an enterprise architecture that blends its data with processes for governance and cybersecurity controls.
00:02:15.199 --> 00:02:17.919
Case study A Nordic Energy Company.
00:02:18.159 --> 00:02:28.240
This energy generation organization operates in a highly regulated sector, managing vast volumes of technical documentation critical to safety in operations.
00:02:28.560 --> 00:02:47.199
Faced with the challenge of enabling over 900 employees to reliably access the latest approved versions of these documents across office and plant environments, the company recognized that legacy, fragmented systems lacked the governance and visibility required for modern risk and trust management.
00:02:47.599 --> 00:02:55.360
In an era where data is both an asset and a vulnerability, establishing a security by design approach became imperative.
00:02:55.599 --> 00:03:08.000
To build a robust foundation, the company implemented a unified content management environment underpinned by strong identity and access controls, workflow automation, and document lifecycle governance.
00:03:08.240 --> 00:03:20.400
By centralizing control and enforcing policy-driven access rights, the system ensured that only the appropriate users could reach sensitive operational records at the right time and from the right context.
00:03:20.639 --> 00:03:33.120
Automated workflows moved documents through review, approval, and archiving in a governed fashion, strengthening the security of the data plane while maintaining usability for field and office teams alike.
00:03:33.280 --> 00:03:45.439
With this architecture in place, the organization laid the groundwork for advanced analytics and AI-enabled capabilities, on the understanding that those must rest on secure, well-managed information.
00:03:45.680 --> 00:03:47.520
The results were transformative.
00:03:47.680 --> 00:03:58.800
The company achieved a high stability record and dramatically improved user productivity with near real-time access to mission-critical content, supporting both safety and operational integrity.
00:03:58.960 --> 00:04:11.280
But perhaps more important, they now possess the trusted information infrastructure required to introduce AI-driven search, insights, and decision support tools, securely and responsibly.
00:04:11.520 --> 00:04:26.560
In short, by treating cybersecurity, data governance, and AI readiness as intertwined components, the company moved from document management to a modern intelligence platform posture, anchored in trust, visibility, and automation.
00:04:26.800 --> 00:04:34.879
We spend upwards of 70% of our time playing defense from a technology perspective, whether it's regulatory or cybersecurity threats.
00:04:35.040 --> 00:04:43.360
We have to remain vigilant in protecting the bank and our clients' data and keep up with the latest changes and patches that addresses vulnerabilities.
00:04:43.839 --> 00:04:46.959
CTO and Managing Director of Global Bank.
00:04:47.120 --> 00:04:50.000
The Cyber Threat Landscape for Data and AI.
00:04:50.240 --> 00:05:12.639
According to the World Economic Forum's Global Cybersecurity Outlook 2025, Gen AI tools are reshaping the cybercrime landscape by enabling criminals to refine their methods and automate and personalize their techniques, with 47% of organizations citing their top concerns surrounding Gen AI as the advance of adversarial capabilities.
00:05:12.800 --> 00:05:19.360
Cyber criminals are harnessing the efficiency of AI to automate and personalize deceptive communications.
00:05:19.600 --> 00:05:30.240
Some 42% of organizations experienced a successful social engineering attack in the past year, a number that can only increase with advances and the malicious adoption of AI.
00:05:30.480 --> 00:05:35.680
Cybersecurity for enterprise AI must be approached from a multidimensional perspective.
00:05:35.839 --> 00:05:42.879
One that recognizes the full spectrum of threats spanning data, models, infrastructure, and human interaction.
00:05:43.120 --> 00:05:49.519
The cyber threat landscape for data and AI runs across infrastructure, governance, and human behavior.
00:05:49.759 --> 00:05:58.720
Traditional cyber risks such as unauthorized access, insider threats, and ransomware, continue to target enterprise systems and compromise critical data.
00:05:58.959 --> 00:06:06.079
And as organizations move more of their data and operations into cloud environments, the overall attack surface continues to grow.
00:06:06.319 --> 00:06:13.120
Threat actors continue to exploit weaknesses in identity management, network segmentation, and vulnerable software.
00:06:13.360 --> 00:06:20.319
These foundational threats create the conditions for more advanced forms of attack that exploit the growing reliance on AI.
00:06:20.480 --> 00:06:27.040
Some of the expanding attack surfaces include the data pipelines used to train LLMs in the models themselves.
00:06:27.279 --> 00:06:40.639
As researchers from IBM and Carnegie Mellon University have observed, growing applications of large language models, LLMs, trained by a third party raise serious concerns on the security vulnerability of LLMs.
00:06:40.879 --> 00:06:50.079
It has been demonstrated that malicious actors can covertly exploit these vulnerabilities in LLMs through poisoning attacks aimed at generating undesirable outputs.
00:06:50.240 --> 00:07:01.600
In addition to model poisoning, other security risks such as data exfiltration and prompt injection are becoming more commonplace, with the latter being one of the biggest challenges for security in LLM.
00:07:01.839 --> 00:07:07.600
Emerging AI-specific threats introduce new vulnerabilities that go beyond conventional data breaches.
00:07:07.759 --> 00:07:14.160
Because new threats emerge every day, it is impossible to build an up-to-date, all-encompassing list of attacks.
00:07:14.399 --> 00:07:28.160
However, some of the most common types include data poisoning attacks, backdoor attacks, adversarial evasion attacks, model inversion attacks, bias exploitation attacks.
00:07:28.720 --> 00:07:35.759
The diagram below shows the AI model lifecycle mapped to the different types of cyber attacks we will review in this section.
00:07:36.000 --> 00:07:42.079
The life cycle begins with data collection and preparation, which feeds into the training phase where the model learns.
00:07:42.399 --> 00:07:54.720
This results in a trained model which is then used for inference, the process of making predictions or decisions, to generate the final output, a prediction, classification, decision, or generated response.
00:07:54.879 --> 00:08:02.000
Understanding how these different cyber attacks relate to the AI model lifecycle puts them in a context for where threats can occur.
00:08:02.160 --> 00:08:04.480
Let's look at the threats in more detail.
00:08:04.879 --> 00:08:05.519
1.
00:08:05.759 --> 00:08:07.120
Data poisoning.
00:08:07.360 --> 00:08:13.680
Data poisoning attacks are common during the data phase in advance of training, through collection and ingestion.
00:08:13.920 --> 00:08:23.040
In these attacks, threat actors inject malicious inputs into the training data set, corrupting how the model learns and undermining its integrity and reliability.
00:08:23.279 --> 00:08:25.920
The root issue lies in a flawed assumption.
00:08:26.160 --> 00:08:31.439
Most learning algorithms presume that training data is clean and representative of reality.
00:08:31.600 --> 00:08:35.919
In security-sensitive environments, that assumption simply doesn't hold true.
00:08:36.240 --> 00:08:36.720
2.
00:08:37.120 --> 00:08:38.720
Backdoor Attacks.
00:08:39.039 --> 00:08:44.399
Backdoor attacks are a form of data poisoning where a trigger pattern is hidden in the model during training.
00:08:44.639 --> 00:08:50.720
The model will behave normally for regular inputs but then will produce a malicious output when the trigger appears.
00:08:50.960 --> 00:08:57.120
These types of attacks can be complicated to detect because they remain inactive until the trigger condition is met.
00:08:57.279 --> 00:09:13.039
In this case, an adversary can create a maliciously trained network, a backdoored neural network, or a bad net, that has state-of-the-art performance on the user's training and validation samples, but behaves badly on specific attacker-chosen inputs.
00:09:13.360 --> 00:09:13.919
3.
00:09:14.240 --> 00:09:19.360
Adversarial Attacks Another common type of attack is an adversarial attack.
00:09:19.519 --> 00:09:25.440
These occur when threat actors try to manipulate the AI model's input to produce incorrect results.
00:09:25.679 --> 00:09:36.720
Sometimes these changes can be so small they aren't recognizable, but they can alter behavior and undermine safety in AI use cases such as medical imaging or autonomous navigation.
00:09:37.039 --> 00:09:37.600
4.
00:09:37.919 --> 00:09:39.759
Model inversion attacks.
00:09:40.080 --> 00:09:43.919
Model inversion attacks pose a threat to privacy and personal data.
00:09:44.159 --> 00:09:51.759
They do this through attempts to reconstruct sensitive input data from model parameters, outputs, or intermediate representations.
00:09:52.000 --> 00:09:59.039
In other words, the attack essentially reverse engineers the model to expose the specific private data it learned from.
00:09:59.360 --> 00:09:59.919
5.
00:10:00.399 --> 00:10:02.559
Bias exploitation attacks.
00:10:02.799 --> 00:10:12.639
The final type of attack we'll highlight is bias exploitation attacks, which take advantage of bias that already exists in the dataset to manipulate decision making.
00:10:12.879 --> 00:10:18.480
These attacks are unlike data poisoning as they do not introduce new data into the dataset.
00:10:18.639 --> 00:10:24.720
Instead, they exploit inherent inequities that are already present in the data to achieve an attack.
00:10:24.960 --> 00:10:37.360
Across both the public and private sectors, risk now extend beyond technical compromise, like gaining system access, to include data manipulation, like altering or poisoning data.
00:10:37.600 --> 00:10:44.399
These are just five examples showing how threat actors are attacking different parts of the AI model lifecycle.
00:10:44.720 --> 00:10:52.960
For example, in the public sector, governments have faced ransomware attacks on public infrastructure that have impacted critical services.
00:10:53.200 --> 00:11:01.440
Likewise, in the private sector, companies have experienced model interference from threat actors that affect their websites and recommendation systems.
00:11:01.759 --> 00:11:05.840
More broadly, Gen AI has been used to spread misinformation.
00:11:06.080 --> 00:11:10.480
These attacks exploit model bias and erode the public trust in AI.
00:11:10.799 --> 00:11:18.159
These cases highlight that cybersecurity for enterprise data in AI is no longer confined to protecting systems.
00:11:18.480 --> 00:11:24.720
It is about defending the integrity of data and decisions and preserving public trust in the cognitive era.
00:11:24.960 --> 00:11:26.799
Data Security Foundations.
00:11:27.039 --> 00:11:30.000
This review of the threats highlight a central theme.
00:11:30.240 --> 00:11:34.480
The data used to train enterprise AI models must be secure.
00:11:34.720 --> 00:11:42.559
As organizations scale their use of AI, the volume and sensitivity of the data they manage will continue to grow exponentially.
00:11:42.799 --> 00:11:49.919
Recall the number of parameters used to train small and large language models and how the volume grows as we get to AGI.
00:11:50.159 --> 00:12:00.000
For enterprises leveraging private data sets to build private AI, it is critical to ensure the confidentiality and integrity of that data throughout its life cycle.
00:12:00.159 --> 00:12:08.559
Building this foundation requires a security by design approach, combining robust technical controls with strong governance mechanisms.
00:12:08.879 --> 00:12:14.720
This strategy is essential to safeguard information while maintaining compliance with standards and regulations.
00:12:15.039 --> 00:12:17.200
Protecting the data lifecycle.
00:12:17.440 --> 00:12:29.200
In general, data goes through different life cycle phases, including collection, storage, transmission, distribution, processing, archiving and retention, and disposition.
00:12:29.600 --> 00:12:38.240
Protecting data at each stage requires a combination of preventative, detective, and corrective controls to defend against cyber threats.
00:12:38.480 --> 00:12:39.360
Collection.
00:12:39.840 --> 00:12:45.440
Data collection needs to be handled carefully as it introduces opportunities for data compromise.
00:12:45.600 --> 00:12:50.000
It is essential to understand what data is being collected and for what purposes.
00:12:50.240 --> 00:13:01.279
ISO slash IEC 27001 2022 provides a framework to help organizations understand how to protect information through its life cycle.
00:13:01.440 --> 00:13:08.559
It offers a set of control categories to ensure that data collection and processing are lawful, fair, and transparent.
00:13:08.879 --> 00:13:09.679
Storage.
00:13:10.000 --> 00:13:14.159
After data has been safely collected, it needs to be securely stored.
00:13:14.320 --> 00:13:18.879
This can be in on-premises infrastructure or in cloud environments.
00:13:19.279 --> 00:13:25.440
General data protection includes encryption at rest, access controls, and segregation of sensitive data.
00:13:25.600 --> 00:13:34.240
It can also include capabilities like immutable storage, which may be part of a broader data protection strategy to mitigate cyber attacks such as ransomware.
00:13:34.399 --> 00:13:38.639
This must be part of your enterprise's zero trust data protection strategy.
00:13:38.879 --> 00:13:40.320
More on this shortly.
00:13:40.639 --> 00:13:46.960
Transmission Data transmitted or distributed between systems is vulnerable to interception.
00:13:47.200 --> 00:13:52.159
To protect this data in transit, technical methodologies like encryption are used.
00:13:52.399 --> 00:13:58.480
Encryption does not prevent data from being intercepted, but it makes the data unusable if it is.
00:13:58.799 --> 00:13:59.759
Processing.
00:14:00.080 --> 00:14:09.840
The data processing stage is a critical point where data can be intercepted or manipulated, making strong access controls essential to prevent unauthorized access.
00:14:10.080 --> 00:14:18.639
At this stage, the primary risk involves privacy breaches, especially when sensitive data sets are used for AI model training or analytics.
00:14:18.799 --> 00:14:22.639
To mitigate these risks, new computation methods have been developed.
00:14:22.879 --> 00:14:30.159
Homomorphic encryption, for example, preserves privacy by enabling operations on encrypted data without decrypting it.
00:14:30.399 --> 00:14:36.000
In addition, federated learning represents a shift towards secure, distributed AI.
00:14:36.320 --> 00:14:41.600
This allows models to be trained locally across multiple decentralized data sets.
00:14:41.919 --> 00:14:51.519
This bring the code to the data approach minimizes the need to centralize sensitive data, reducing exposure risk while maintaining model performance.
00:14:51.759 --> 00:14:53.919
Data disposal or deletion.
00:14:54.480 --> 00:15:01.279
Secure deletion of data is the final stage and ensures that old or redundant data is permanently removed.
00:15:01.600 --> 00:15:15.600
Under privacy regulations, such as GDPR Article 17, there is a right to eraser, or right to be forgotten, which means that organizations must demonstrate that they have effectively executed deletion requests.
00:15:15.840 --> 00:15:22.399
Each data lifecycle stage is interdependent, and a weakness in any one can compromise the entire life cycle.
00:15:22.639 --> 00:15:31.279
Understanding the risk across the life cycle ensures that, as you develop a zero trust data protection strategy, you have considered all aspects.
00:15:31.519 --> 00:15:44.720
In the following case study, a leading chemicals company is using an enterprise information management system to manage their data lifecycle, achieve compliance, and secure their data across multiple processes, partners, and locations.
00:15:45.440 --> 00:15:47.519
Case study Lanxus.
00:15:48.080 --> 00:15:57.039
The core business of Lanxis is the development, manufacturing, and marketing of chemical intermediates, additives, specialty chemicals, and plastics.
00:15:57.279 --> 00:16:02.320
What follows are excerpts of an interview with a company's process expert, ECM.
00:16:02.720 --> 00:16:16.000
Given the complexity of our portfolio, when we manufacture products, chemical intermediates, additives, specialty chemicals, and plastics, our processes leave paper trails that stem from scientific research through to sales and marketing.
00:16:16.240 --> 00:16:21.039
A typical starting point for research occurs when a customer requests a new product feature.
00:16:21.200 --> 00:16:28.480
Typically, we would then conduct research with an external partner, so there would be requirements around secure access and collaboration.
00:16:28.720 --> 00:16:36.960
Because we are a company that manufactures and distributes globally, our products, operations, and paper trails have to comply with global regulations.
00:16:37.279 --> 00:16:53.519
An Enterprise Content Management ECM platform helps us to ensure that information is compliant, from the research conducted to the procedures that engineers establish to manufacture at scale, to the construction and operation of a plant, and finally through to sales and marketing.
00:16:53.679 --> 00:16:56.399
We deal with large volumes of paper every day.
00:16:56.559 --> 00:17:05.519
Every step in a process has to be compliant and well documented, especially given that we operate in 25 countries and each one has a different set of regulations.
00:17:05.839 --> 00:17:15.839
Compliance is a benefit as a result of effective information management, along with efficiency and productivity, specifically being able to find information more quickly.
00:17:16.079 --> 00:17:22.640
To realize these benefits, we have to show our internal customers how using the technology will make their jobs easier.
00:17:22.880 --> 00:17:27.759
ECM delivers the tools we need to balance compliance and security with usability.
00:17:28.079 --> 00:17:31.039
Zero Trust Architecture for Enterprise AI.
00:17:31.200 --> 00:17:39.039
We have reviewed cyber threats related to data in AI, focusing specifically on the data lifecycle and the points where attacks can occur.
00:17:39.279 --> 00:17:54.240
To protect against these threats, the National Institute of Standards and Technology, NIST, defines a zero trust architecture, ZTA, as a strategic approach to cybersecurity that assumes no implicit trust within a network.
00:17:54.559 --> 00:17:59.200
The model is based on the idea that you should never trust and always verify.
00:17:59.359 --> 00:18:02.480
And this philosophy must govern every access decision.
00:18:02.640 --> 00:18:13.839
Instead of relying on defenses such as firewalls or VPNs, Zero Trust leverages continuous verification and access controls across all assets, users, and data flows.
00:18:14.000 --> 00:18:31.200
According to NIST SP 800-207, the Zero Trust model redefines traditional enterprise security by focusing on identity-centric protection, every access request must be authenticated and authorized in real time.
00:18:31.519 --> 00:18:33.279
Least privilege access.
00:18:33.519 --> 00:18:38.960
Users and systems receive only the minimum level of access necessary to perform their function.
00:18:39.200 --> 00:18:41.200
Dynamic policy enforcement.
00:18:41.440 --> 00:18:48.079
Access decisions are evaluated in real time based on factors such as user behavior and data sensitivity.
00:18:48.319 --> 00:18:49.680
Micro segmentation.
00:18:49.920 --> 00:18:56.720
Networks are divided into small, isolated zones to limit the movement of threat actors in the event of a compromise.
00:18:57.440 --> 00:18:59.039
Visibility and analytics.
00:18:59.440 --> 00:19:05.440
Continuous monitoring and threat detection ensure that certain behaviors trigger automated responses.
00:19:05.680 --> 00:19:08.160
Zero trust is not a single solution.
00:19:08.319 --> 00:19:22.799
It is achieved through a combination of technology solutions, including identity and access management, multifactor authentication, MFA, encryption, continuous monitoring, and automated policy enforcement.
00:19:23.119 --> 00:19:28.480
AIAM will become a critical security component of any enterprise system.
00:19:28.799 --> 00:19:38.799
Find out how a Latin American entertainment company is working to combine these technology solutions in their plan to evolve to a zero trust security model in the feature below.
00:19:39.119 --> 00:19:43.039
Case study A Latin American Entertainment Company.
00:19:44.079 --> 00:19:56.480
With millions of customers and thousands of employees across multiple countries, a leading Latin American entertainment company faced growing challenges managing identity and access for a large, distributed workforce.
00:19:57.119 --> 00:20:07.119
Over time, fragmented systems and manual provisioning made it difficult to maintain visibility across 15,000 user identities and more than 400 applications.
00:20:07.359 --> 00:20:15.680
The lack of unified governance slowed response times, created security blind spots, and made it harder to advance toward a zero trust model.
00:20:15.920 --> 00:20:28.559
To address this, the company implemented a comprehensive identity governance and administration IGA framework, consolidating global identity data into a single source of truth and central point of control.
00:20:28.799 --> 00:20:41.599
Integrated with HR Systems, Active Directory, and dozens of enterprise applications, the platform automated provisioning, deprovisioning, and access reviews, reducing manual workloads by half.
00:20:41.920 --> 00:20:53.759
Intelligent alerting, continuous attestation, and role-based access controls reinforced compliance, minimized risk, and enforced least privilege principles across the enterprise.
00:20:54.079 --> 00:20:55.920
The results were immediate.
00:20:56.160 --> 00:21:07.039
The company gained end-to-end visibility into more than 15,000 identities, streamlined access management, and strengthened its security posture across a global footprint.
00:21:07.279 --> 00:21:23.119
With identity governance now at the core of its cybersecurity strategy, the company is well positioned to advance its zero trust model, extending the same rigor and automation to protect its data, applications, and AI-driven operations across the digital enterprise.
00:21:23.440 --> 00:21:25.839
AI security and model protection.
00:21:26.079 --> 00:21:31.599
Zero trust, as we just reviewed, provides a philosophy and strategy for protecting access.
00:21:31.759 --> 00:21:36.160
But we must also consider AI security and model protection more broadly.
00:21:36.400 --> 00:21:43.759
AI models differ from traditional IT systems in that they combine logic with the ability to learn from data continuously.
00:21:44.000 --> 00:21:48.319
We reviewed earlier the attack surfaces throughout the AI model lifecycle.
00:21:48.480 --> 00:21:56.160
To protect against these risks, organizations are adopting strategies that combine classic cybersecurity approaches with new ones.
00:21:56.400 --> 00:21:59.599
This can include training teams and models on adversarial attack.
00:22:00.240 --> 00:22:06.559
Approaches, watermarking models, and running red team tests of their own pre-deployment environments.
00:22:06.880 --> 00:22:13.759
For context, a red team is a group that simulates real-world cyber attacks to test an organization's security.
00:22:13.920 --> 00:22:18.079
Their goal is to find weaknesses in systems, networks, and people.
00:22:18.799 --> 00:22:27.680
Training can improve model performance by exposing it to adversarial examples during training, thereby increasing its resilience against adversarial inputs.
00:22:27.920 --> 00:22:33.359
Model watermarking provides assurance and helps identify unauthorized reuse of models.
00:22:33.599 --> 00:22:39.039
Red team tests are helpful in exposing vulnerabilities through simulated attacks before deployment.
00:22:39.279 --> 00:22:44.880
In combination with a zero trust approach, these can be powerful tactics to protect against attacks.
00:22:45.039 --> 00:22:53.759
However, these are just a few of the many potential approaches, and this should be defined as part of an overall AI strategy across the enterprise.
00:22:54.000 --> 00:22:57.680
Looking ahead, the future of cybersecurity for AI.
00:22:57.920 --> 00:23:07.839
This chapter has looked at the growing importance of cybersecurity in relation to enterprise AI, highlighting the rising number of cyber attacks targeting enterprise AI systems.
00:23:08.079 --> 00:23:16.559
With reports indicating that 62% of organizations have faced deep fake attacks and concerns about Gen AI's adversarial capabilities.
00:23:16.799 --> 00:23:21.279
The urgency of addressing AI and data-related cyber risk is apparent.
00:23:22.000 --> 00:23:32.880
As organizations leverage private data to enhance operational efficiency, they are simultaneously exposing themselves to complex vulnerabilities that threaten their AI models and data.
00:23:33.200 --> 00:23:41.519
We also analyze how these attacks work, outlining threats like data poisoning, backdoor attacks, and model inversion attacks.
00:23:41.759 --> 00:23:48.880
These risks illuminate some of the limitations of traditional cybersecurity approaches in protecting advanced AI systems.
00:23:49.119 --> 00:24:01.680
By understanding how these attacks relate to the different phases of the AI model lifecycle, organizations can better anticipate potential vulnerabilities and implement strategies for data protection and model security.
00:24:02.000 --> 00:24:12.799
Looking to the future, organizations must adopt proactive, adaptive cybersecurity frameworks that incorporate AI-driven defense approaches to counter AI-powered attacks.
00:24:13.039 --> 00:24:18.480
This includes developing intelligent threat detection systems and new risk assessment models.
00:24:18.640 --> 00:24:33.680
Ultimately, collaboration across public and private sectors, coupled with investment in innovative cybersecurity solutions, will be essential to outpace the evolving threat landscape and ensure the safe integration of AI technologies into enterprise operations.
00:24:33.920 --> 00:24:38.799
As organizations strengthen their cyber defenses, one truth becomes clear.
00:24:39.119 --> 00:24:41.920
Security and trust are inseparable.
00:24:42.160 --> 00:24:47.039
Protecting enterprise AI systems isn't only about defending against attacks.
00:24:47.200 --> 00:24:54.160
It's about ensuring that the data powering those systems remains accurate, ethical, and reliable.
00:24:54.480 --> 00:24:59.839
In the next chapter, we'll explore the foundation of trusted AI, data governance.
00:25:00.240 --> 00:25:02.240
The Fast Five Download.
00:25:02.559 --> 00:25:08.000
One, adopt a zero trust architecture for all data in AI systems.
00:25:08.319 --> 00:25:14.240
Immediately implement a zero trust security model that assumes no implicit trust within your network.
00:25:14.480 --> 00:25:26.480
Enforce continuous identity verification, least privilege access, dynamic policy enforcement, and micro segmentation to minimize the risk of both internal and external breaches.
00:25:26.799 --> 00:25:27.279
2.
00:25:27.839 --> 00:25:31.519
Secure the entire data lifecycle with integrated controls.
00:25:31.839 --> 00:25:40.799
Mandate that all data across collection, storage, transmission, processing, and disposal is protected with layered security measures.
00:25:41.039 --> 00:26:01.759
This includes encryption at rest and in transit, strong access controls, immutable storage, and strict compliance with regulations, like GDPR and ISO slash IEC 27001 2022, any gaps in one stage can compromise the entire system.
00:26:02.079 --> 00:26:02.720
3.
00:26:03.119 --> 00:26:06.079
Harden AI models against emerging threats.
00:26:06.319 --> 00:26:17.440
Establish protocols to defend against AI-specific attacks such as data poisoning, backdoor exploits, adversarial inputs, model inversion, and bias exploitation.
00:26:17.599 --> 00:26:26.880
Incorporate adversarial training, model watermarking, and regular red team testing to identify and remediate vulnerabilities before deployment.
00:26:27.200 --> 00:26:31.839
Four, build security by design into AI initiatives.
00:26:32.400 --> 00:26:39.039
Insist that every new AI or data project incorporates security and privacy by design from inception.
00:26:39.279 --> 00:26:52.079
Require cross-functional teams, including data, IT, compliance, and security, to work together to ensure technical and governance controls are integrated into AI model development and operations.
00:26:52.640 --> 00:26:58.640
Five, invest in proactive, enterprise AI-driven cybersecurity capabilities.
00:26:58.960 --> 00:27:05.839
Allocate resources to develop and deploy intelligent, adaptive cybersecurity solutions powered by AI.
00:27:06.160 --> 00:27:15.839
These should include automated threat detection, risk assessment tools, and real time monitoring to keep pace with evolving AI enabled attack methods.
00:27:16.079 --> 00:27:21.599
Foster collaboration with industry peers and public sector partners to stay ahead of emerging threats.
00:00:01.919 --> 00:00:14.160
Chapter four Making It Secure The Importance of Cybersecurity Innovation must be balanced by trust to ensure that the intelligence we build cannot be turned against us.
00:00:14.400 --> 00:00:19.679
In this chapter, we explore how cybersecurity must evolve alongside AI.
00:00:19.920 --> 00:00:26.320
As intelligence systems reshape how enterprises operate, new risks demand equally advanced defenses.
00:00:26.480 --> 00:00:34.399
We'll examine emerging threats and the strategies required to secure data, models, and AI-driven operations against them.
00:00:34.719 --> 00:00:50.799
62% of organizations experienced a deep fake attack involving social engineering or exploiting automated processes, while 32% said they experienced an attack on AI applications that leveraged the application prompt in the last 12 months.
00:00:51.119 --> 00:00:59.280
In recent years, cyber threats have evolved from simple breaches to sophisticated assaults aimed directly at enterprise AI systems.
00:00:59.439 --> 00:01:01.679
And the stakes have never been higher.
00:01:01.920 --> 00:01:15.120
As organizations accelerate adoption of technologies such as generative AI, Gen AI, we are witnessing a surge in attacks that leverage AI for phishing, deepfakes, and advanced social engineering.
00:01:15.359 --> 00:01:30.000
At the same time, a new wave of vulnerability is emerging: malicious actors who exploit Gen AI infrastructure, manipulate prompts, or compromise chained AI workflows to infiltrate and disrupt enterprises.
00:01:30.319 --> 00:01:38.799
In the previous chapter, we examined how the intersection of data and enterprise AI creates opportunities for innovation and operational efficiency.
00:01:39.040 --> 00:01:42.079
But opportunities also bring unwanted risks.
00:01:42.239 --> 00:01:51.359
And as organizations rely on private data to fuel their AI engines, they may inadvertently expose themselves to new and evolving cyber risks.
00:01:51.599 --> 00:02:00.480
The protection of enterprise data and AI must keep pace with technological evolution, as data and AI are attractive targets for threat actors.
00:02:00.719 --> 00:02:14.960
In the following feature, an energy company lays the foundation for AI and advanced analytics within a secure EIM system, building an enterprise architecture that blends its data with processes for governance and cybersecurity controls.
00:02:15.199 --> 00:02:17.919
Case study A Nordic Energy Company.
00:02:18.159 --> 00:02:28.240
This energy generation organization operates in a highly regulated sector, managing vast volumes of technical documentation critical to safety in operations.
00:02:28.560 --> 00:02:47.199
Faced with the challenge of enabling over 900 employees to reliably access the latest approved versions of these documents across office and plant environments, the company recognized that legacy, fragmented systems lacked the governance and visibility required for modern risk and trust management.
00:02:47.599 --> 00:02:55.360
In an era where data is both an asset and a vulnerability, establishing a security by design approach became imperative.
00:02:55.599 --> 00:03:08.000
To build a robust foundation, the company implemented a unified content management environment underpinned by strong identity and access controls, workflow automation, and document lifecycle governance.
00:03:08.240 --> 00:03:20.400
By centralizing control and enforcing policy-driven access rights, the system ensured that only the appropriate users could reach sensitive operational records at the right time and from the right context.
00:03:20.639 --> 00:03:33.120
Automated workflows moved documents through review, approval, and archiving in a governed fashion, strengthening the security of the data plane while maintaining usability for field and office teams alike.
00:03:33.280 --> 00:03:45.439
With this architecture in place, the organization laid the groundwork for advanced analytics and AI-enabled capabilities, on the understanding that those must rest on secure, well-managed information.
00:03:45.680 --> 00:03:47.520
The results were transformative.
00:03:47.680 --> 00:03:58.800
The company achieved a high stability record and dramatically improved user productivity with near real-time access to mission-critical content, supporting both safety and operational integrity.
00:03:58.960 --> 00:04:11.280
But perhaps more important, they now possess the trusted information infrastructure required to introduce AI-driven search, insights, and decision support tools, securely and responsibly.
00:04:11.520 --> 00:04:26.560
In short, by treating cybersecurity, data governance, and AI readiness as intertwined components, the company moved from document management to a modern intelligence platform posture, anchored in trust, visibility, and automation.
00:04:26.800 --> 00:04:34.879
We spend upwards of 70% of our time playing defense from a technology perspective, whether it's regulatory or cybersecurity threats.
00:04:35.040 --> 00:04:43.360
We have to remain vigilant in protecting the bank and our clients' data and keep up with the latest changes and patches that addresses vulnerabilities.
00:04:43.839 --> 00:04:46.959
CTO and Managing Director of Global Bank.
00:04:47.120 --> 00:04:50.000
The Cyber Threat Landscape for Data and AI.
00:04:50.240 --> 00:05:12.639
According to the World Economic Forum's Global Cybersecurity Outlook 2025, Gen AI tools are reshaping the cybercrime landscape by enabling criminals to refine their methods and automate and personalize their techniques, with 47% of organizations citing their top concerns surrounding Gen AI as the advance of adversarial capabilities.
00:05:12.800 --> 00:05:19.360
Cyber criminals are harnessing the efficiency of AI to automate and personalize deceptive communications.
00:05:19.600 --> 00:05:30.240
Some 42% of organizations experienced a successful social engineering attack in the past year, a number that can only increase with advances and the malicious adoption of AI.
00:05:30.480 --> 00:05:35.680
Cybersecurity for enterprise AI must be approached from a multidimensional perspective.
00:05:35.839 --> 00:05:42.879
One that recognizes the full spectrum of threats spanning data, models, infrastructure, and human interaction.
00:05:43.120 --> 00:05:49.519
The cyber threat landscape for data and AI runs across infrastructure, governance, and human behavior.
00:05:49.759 --> 00:05:58.720
Traditional cyber risks such as unauthorized access, insider threats, and ransomware, continue to target enterprise systems and compromise critical data.
00:05:58.959 --> 00:06:06.079
And as organizations move more of their data and operations into cloud environments, the overall attack surface continues to grow.
00:06:06.319 --> 00:06:13.120
Threat actors continue to exploit weaknesses in identity management, network segmentation, and vulnerable software.
00:06:13.360 --> 00:06:20.319
These foundational threats create the conditions for more advanced forms of attack that exploit the growing reliance on AI.
00:06:20.480 --> 00:06:27.040
Some of the expanding attack surfaces include the data pipelines used to train LLMs in the models themselves.
00:06:27.279 --> 00:06:40.639
As researchers from IBM and Carnegie Mellon University have observed, growing applications of large language models, LLMs, trained by a third party raise serious concerns on the security vulnerability of LLMs.
00:06:40.879 --> 00:06:50.079
It has been demonstrated that malicious actors can covertly exploit these vulnerabilities in LLMs through poisoning attacks aimed at generating undesirable outputs.
00:06:50.240 --> 00:07:01.600
In addition to model poisoning, other security risks such as data exfiltration and prompt injection are becoming more commonplace, with the latter being one of the biggest challenges for security in LLM.
00:07:01.839 --> 00:07:07.600
Emerging AI-specific threats introduce new vulnerabilities that go beyond conventional data breaches.
00:07:07.759 --> 00:07:14.160
Because new threats emerge every day, it is impossible to build an up-to-date, all-encompassing list of attacks.
00:07:14.399 --> 00:07:28.160
However, some of the most common types include data poisoning attacks, backdoor attacks, adversarial evasion attacks, model inversion attacks, bias exploitation attacks.
00:07:28.720 --> 00:07:35.759
The diagram below shows the AI model lifecycle mapped to the different types of cyber attacks we will review in this section.
00:07:36.000 --> 00:07:42.079
The life cycle begins with data collection and preparation, which feeds into the training phase where the model learns.
00:07:42.399 --> 00:07:54.720
This results in a trained model which is then used for inference, the process of making predictions or decisions, to generate the final output, a prediction, classification, decision, or generated response.
00:07:54.879 --> 00:08:02.000
Understanding how these different cyber attacks relate to the AI model lifecycle puts them in a context for where threats can occur.
00:08:02.160 --> 00:08:04.480
Let's look at the threats in more detail.
00:08:04.879 --> 00:08:05.519
1.
00:08:05.759 --> 00:08:07.120
Data poisoning.
00:08:07.360 --> 00:08:13.680
Data poisoning attacks are common during the data phase in advance of training, through collection and ingestion.
00:08:13.920 --> 00:08:23.040
In these attacks, threat actors inject malicious inputs into the training data set, corrupting how the model learns and undermining its integrity and reliability.
00:08:23.279 --> 00:08:25.920
The root issue lies in a flawed assumption.
00:08:26.160 --> 00:08:31.439
Most learning algorithms presume that training data is clean and representative of reality.
00:08:31.600 --> 00:08:35.919
In security-sensitive environments, that assumption simply doesn't hold true.
00:08:36.240 --> 00:08:36.720
2.
00:08:37.120 --> 00:08:38.720
Backdoor Attacks.
00:08:39.039 --> 00:08:44.399
Backdoor attacks are a form of data poisoning where a trigger pattern is hidden in the model during training.
00:08:44.639 --> 00:08:50.720
The model will behave normally for regular inputs but then will produce a malicious output when the trigger appears.
00:08:50.960 --> 00:08:57.120
These types of attacks can be complicated to detect because they remain inactive until the trigger condition is met.
00:08:57.279 --> 00:09:13.039
In this case, an adversary can create a maliciously trained network, a backdoored neural network, or a bad net, that has state-of-the-art performance on the user's training and validation samples, but behaves badly on specific attacker-chosen inputs.
00:09:13.360 --> 00:09:13.919
3.
00:09:14.240 --> 00:09:19.360
Adversarial Attacks Another common type of attack is an adversarial attack.
00:09:19.519 --> 00:09:25.440
These occur when threat actors try to manipulate the AI model's input to produce incorrect results.
00:09:25.679 --> 00:09:36.720
Sometimes these changes can be so small they aren't recognizable, but they can alter behavior and undermine safety in AI use cases such as medical imaging or autonomous navigation.
00:09:37.039 --> 00:09:37.600
4.
00:09:37.919 --> 00:09:39.759
Model inversion attacks.
00:09:40.080 --> 00:09:43.919
Model inversion attacks pose a threat to privacy and personal data.
00:09:44.159 --> 00:09:51.759
They do this through attempts to reconstruct sensitive input data from model parameters, outputs, or intermediate representations.
00:09:52.000 --> 00:09:59.039
In other words, the attack essentially reverse engineers the model to expose the specific private data it learned from.
00:09:59.360 --> 00:09:59.919
5.
00:10:00.399 --> 00:10:02.559
Bias exploitation attacks.
00:10:02.799 --> 00:10:12.639
The final type of attack we'll highlight is bias exploitation attacks, which take advantage of bias that already exists in the dataset to manipulate decision making.
00:10:12.879 --> 00:10:18.480
These attacks are unlike data poisoning as they do not introduce new data into the dataset.
00:10:18.639 --> 00:10:24.720
Instead, they exploit inherent inequities that are already present in the data to achieve an attack.
00:10:24.960 --> 00:10:37.360
Across both the public and private sectors, risk now extend beyond technical compromise, like gaining system access, to include data manipulation, like altering or poisoning data.
00:10:37.600 --> 00:10:44.399
These are just five examples showing how threat actors are attacking different parts of the AI model lifecycle.
00:10:44.720 --> 00:10:52.960
For example, in the public sector, governments have faced ransomware attacks on public infrastructure that have impacted critical services.
00:10:53.200 --> 00:11:01.440
Likewise, in the private sector, companies have experienced model interference from threat actors that affect their websites and recommendation systems.
00:11:01.759 --> 00:11:05.840
More broadly, Gen AI has been used to spread misinformation.
00:11:06.080 --> 00:11:10.480
These attacks exploit model bias and erode the public trust in AI.
00:11:10.799 --> 00:11:18.159
These cases highlight that cybersecurity for enterprise data in AI is no longer confined to protecting systems.
00:11:18.480 --> 00:11:24.720
It is about defending the integrity of data and decisions and preserving public trust in the cognitive era.
00:11:24.960 --> 00:11:26.799
Data Security Foundations.
00:11:27.039 --> 00:11:30.000
This review of the threats highlight a central theme.
00:11:30.240 --> 00:11:34.480
The data used to train enterprise AI models must be secure.
00:11:34.720 --> 00:11:42.559
As organizations scale their use of AI, the volume and sensitivity of the data they manage will continue to grow exponentially.
00:11:42.799 --> 00:11:49.919
Recall the number of parameters used to train small and large language models and how the volume grows as we get to AGI.
00:11:50.159 --> 00:12:00.000
For enterprises leveraging private data sets to build private AI, it is critical to ensure the confidentiality and integrity of that data throughout its life cycle.
00:12:00.159 --> 00:12:08.559
Building this foundation requires a security by design approach, combining robust technical controls with strong governance mechanisms.
00:12:08.879 --> 00:12:14.720
This strategy is essential to safeguard information while maintaining compliance with standards and regulations.
00:12:15.039 --> 00:12:17.200
Protecting the data lifecycle.
00:12:17.440 --> 00:12:29.200
In general, data goes through different life cycle phases, including collection, storage, transmission, distribution, processing, archiving and retention, and disposition.
00:12:29.600 --> 00:12:38.240
Protecting data at each stage requires a combination of preventative, detective, and corrective controls to defend against cyber threats.
00:12:38.480 --> 00:12:39.360
Collection.
00:12:39.840 --> 00:12:45.440
Data collection needs to be handled carefully as it introduces opportunities for data compromise.
00:12:45.600 --> 00:12:50.000
It is essential to understand what data is being collected and for what purposes.
00:12:50.240 --> 00:13:01.279
ISO slash IEC 27001 2022 provides a framework to help organizations understand how to protect information through its life cycle.
00:13:01.440 --> 00:13:08.559
It offers a set of control categories to ensure that data collection and processing are lawful, fair, and transparent.
00:13:08.879 --> 00:13:09.679
Storage.
00:13:10.000 --> 00:13:14.159
After data has been safely collected, it needs to be securely stored.
00:13:14.320 --> 00:13:18.879
This can be in on-premises infrastructure or in cloud environments.
00:13:19.279 --> 00:13:25.440
General data protection includes encryption at rest, access controls, and segregation of sensitive data.
00:13:25.600 --> 00:13:34.240
It can also include capabilities like immutable storage, which may be part of a broader data protection strategy to mitigate cyber attacks such as ransomware.
00:13:34.399 --> 00:13:38.639
This must be part of your enterprise's zero trust data protection strategy.
00:13:38.879 --> 00:13:40.320
More on this shortly.
00:13:40.639 --> 00:13:46.960
Transmission Data transmitted or distributed between systems is vulnerable to interception.
00:13:47.200 --> 00:13:52.159
To protect this data in transit, technical methodologies like encryption are used.
00:13:52.399 --> 00:13:58.480
Encryption does not prevent data from being intercepted, but it makes the data unusable if it is.
00:13:58.799 --> 00:13:59.759
Processing.
00:14:00.080 --> 00:14:09.840
The data processing stage is a critical point where data can be intercepted or manipulated, making strong access controls essential to prevent unauthorized access.
00:14:10.080 --> 00:14:18.639
At this stage, the primary risk involves privacy breaches, especially when sensitive data sets are used for AI model training or analytics.
00:14:18.799 --> 00:14:22.639
To mitigate these risks, new computation methods have been developed.
00:14:22.879 --> 00:14:30.159
Homomorphic encryption, for example, preserves privacy by enabling operations on encrypted data without decrypting it.
00:14:30.399 --> 00:14:36.000
In addition, federated learning represents a shift towards secure, distributed AI.
00:14:36.320 --> 00:14:41.600
This allows models to be trained locally across multiple decentralized data sets.
00:14:41.919 --> 00:14:51.519
This bring the code to the data approach minimizes the need to centralize sensitive data, reducing exposure risk while maintaining model performance.
00:14:51.759 --> 00:14:53.919
Data disposal or deletion.
00:14:54.480 --> 00:15:01.279
Secure deletion of data is the final stage and ensures that old or redundant data is permanently removed.
00:15:01.600 --> 00:15:15.600
Under privacy regulations, such as GDPR Article 17, there is a right to eraser, or right to be forgotten, which means that organizations must demonstrate that they have effectively executed deletion requests.
00:15:15.840 --> 00:15:22.399
Each data lifecycle stage is interdependent, and a weakness in any one can compromise the entire life cycle.
00:15:22.639 --> 00:15:31.279
Understanding the risk across the life cycle ensures that, as you develop a zero trust data protection strategy, you have considered all aspects.
00:15:31.519 --> 00:15:44.720
In the following case study, a leading chemicals company is using an enterprise information management system to manage their data lifecycle, achieve compliance, and secure their data across multiple processes, partners, and locations.
00:15:45.440 --> 00:15:47.519
Case study Lanxus.
00:15:48.080 --> 00:15:57.039
The core business of Lanxis is the development, manufacturing, and marketing of chemical intermediates, additives, specialty chemicals, and plastics.
00:15:57.279 --> 00:16:02.320
What follows are excerpts of an interview with a company's process expert, ECM.
00:16:02.720 --> 00:16:16.000
Given the complexity of our portfolio, when we manufacture products, chemical intermediates, additives, specialty chemicals, and plastics, our processes leave paper trails that stem from scientific research through to sales and marketing.
00:16:16.240 --> 00:16:21.039
A typical starting point for research occurs when a customer requests a new product feature.
00:16:21.200 --> 00:16:28.480
Typically, we would then conduct research with an external partner, so there would be requirements around secure access and collaboration.
00:16:28.720 --> 00:16:36.960
Because we are a company that manufactures and distributes globally, our products, operations, and paper trails have to comply with global regulations.
00:16:37.279 --> 00:16:53.519
An Enterprise Content Management ECM platform helps us to ensure that information is compliant, from the research conducted to the procedures that engineers establish to manufacture at scale, to the construction and operation of a plant, and finally through to sales and marketing.
00:16:53.679 --> 00:16:56.399
We deal with large volumes of paper every day.
00:16:56.559 --> 00:17:05.519
Every step in a process has to be compliant and well documented, especially given that we operate in 25 countries and each one has a different set of regulations.
00:17:05.839 --> 00:17:15.839
Compliance is a benefit as a result of effective information management, along with efficiency and productivity, specifically being able to find information more quickly.
00:17:16.079 --> 00:17:22.640
To realize these benefits, we have to show our internal customers how using the technology will make their jobs easier.
00:17:22.880 --> 00:17:27.759
ECM delivers the tools we need to balance compliance and security with usability.
00:17:28.079 --> 00:17:31.039
Zero Trust Architecture for Enterprise AI.
00:17:31.200 --> 00:17:39.039
We have reviewed cyber threats related to data in AI, focusing specifically on the data lifecycle and the points where attacks can occur.
00:17:39.279 --> 00:17:54.240
To protect against these threats, the National Institute of Standards and Technology, NIST, defines a zero trust architecture, ZTA, as a strategic approach to cybersecurity that assumes no implicit trust within a network.
00:17:54.559 --> 00:17:59.200
The model is based on the idea that you should never trust and always verify.
00:17:59.359 --> 00:18:02.480
And this philosophy must govern every access decision.
00:18:02.640 --> 00:18:13.839
Instead of relying on defenses such as firewalls or VPNs, Zero Trust leverages continuous verification and access controls across all assets, users, and data flows.
00:18:14.000 --> 00:18:31.200
According to NIST SP 800-207, the Zero Trust model redefines traditional enterprise security by focusing on identity-centric protection, every access request must be authenticated and authorized in real time.
00:18:31.519 --> 00:18:33.279
Least privilege access.
00:18:33.519 --> 00:18:38.960
Users and systems receive only the minimum level of access necessary to perform their function.
00:18:39.200 --> 00:18:41.200
Dynamic policy enforcement.
00:18:41.440 --> 00:18:48.079
Access decisions are evaluated in real time based on factors such as user behavior and data sensitivity.
00:18:48.319 --> 00:18:49.680
Micro segmentation.
00:18:49.920 --> 00:18:56.720
Networks are divided into small, isolated zones to limit the movement of threat actors in the event of a compromise.
00:18:57.440 --> 00:18:59.039
Visibility and analytics.
00:18:59.440 --> 00:19:05.440
Continuous monitoring and threat detection ensure that certain behaviors trigger automated responses.
00:19:05.680 --> 00:19:08.160
Zero trust is not a single solution.
00:19:08.319 --> 00:19:22.799
It is achieved through a combination of technology solutions, including identity and access management, multifactor authentication, MFA, encryption, continuous monitoring, and automated policy enforcement.
00:19:23.119 --> 00:19:28.480
AIAM will become a critical security component of any enterprise system.
00:19:28.799 --> 00:19:38.799
Find out how a Latin American entertainment company is working to combine these technology solutions in their plan to evolve to a zero trust security model in the feature below.
00:19:39.119 --> 00:19:43.039
Case study A Latin American Entertainment Company.
00:19:44.079 --> 00:19:56.480
With millions of customers and thousands of employees across multiple countries, a leading Latin American entertainment company faced growing challenges managing identity and access for a large, distributed workforce.
00:19:57.119 --> 00:20:07.119
Over time, fragmented systems and manual provisioning made it difficult to maintain visibility across 15,000 user identities and more than 400 applications.
00:20:07.359 --> 00:20:15.680
The lack of unified governance slowed response times, created security blind spots, and made it harder to advance toward a zero trust model.
00:20:15.920 --> 00:20:28.559
To address this, the company implemented a comprehensive identity governance and administration IGA framework, consolidating global identity data into a single source of truth and central point of control.
00:20:28.799 --> 00:20:41.599
Integrated with HR Systems, Active Directory, and dozens of enterprise applications, the platform automated provisioning, deprovisioning, and access reviews, reducing manual workloads by half.
00:20:41.920 --> 00:20:53.759
Intelligent alerting, continuous attestation, and role-based access controls reinforced compliance, minimized risk, and enforced least privilege principles across the enterprise.
00:20:54.079 --> 00:20:55.920
The results were immediate.
00:20:56.160 --> 00:21:07.039
The company gained end-to-end visibility into more than 15,000 identities, streamlined access management, and strengthened its security posture across a global footprint.
00:21:07.279 --> 00:21:23.119
With identity governance now at the core of its cybersecurity strategy, the company is well positioned to advance its zero trust model, extending the same rigor and automation to protect its data, applications, and AI-driven operations across the digital enterprise.
00:21:23.440 --> 00:21:25.839
AI security and model protection.
00:21:26.079 --> 00:21:31.599
Zero trust, as we just reviewed, provides a philosophy and strategy for protecting access.
00:21:31.759 --> 00:21:36.160
But we must also consider AI security and model protection more broadly.
00:21:36.400 --> 00:21:43.759
AI models differ from traditional IT systems in that they combine logic with the ability to learn from data continuously.
00:21:44.000 --> 00:21:48.319
We reviewed earlier the attack surfaces throughout the AI model lifecycle.
00:21:48.480 --> 00:21:56.160
To protect against these risks, organizations are adopting strategies that combine classic cybersecurity approaches with new ones.
00:21:56.400 --> 00:21:59.599
This can include training teams and models on adversarial attack.
00:22:00.240 --> 00:22:06.559
Approaches, watermarking models, and running red team tests of their own pre-deployment environments.
00:22:06.880 --> 00:22:13.759
For context, a red team is a group that simulates real-world cyber attacks to test an organization's security.
00:22:13.920 --> 00:22:18.079
Their goal is to find weaknesses in systems, networks, and people.
00:22:18.799 --> 00:22:27.680
Training can improve model performance by exposing it to adversarial examples during training, thereby increasing its resilience against adversarial inputs.
00:22:27.920 --> 00:22:33.359
Model watermarking provides assurance and helps identify unauthorized reuse of models.
00:22:33.599 --> 00:22:39.039
Red team tests are helpful in exposing vulnerabilities through simulated attacks before deployment.
00:22:39.279 --> 00:22:44.880
In combination with a zero trust approach, these can be powerful tactics to protect against attacks.
00:22:45.039 --> 00:22:53.759
However, these are just a few of the many potential approaches, and this should be defined as part of an overall AI strategy across the enterprise.
00:22:54.000 --> 00:22:57.680
Looking ahead, the future of cybersecurity for AI.
00:22:57.920 --> 00:23:07.839
This chapter has looked at the growing importance of cybersecurity in relation to enterprise AI, highlighting the rising number of cyber attacks targeting enterprise AI systems.
00:23:08.079 --> 00:23:16.559
With reports indicating that 62% of organizations have faced deep fake attacks and concerns about Gen AI's adversarial capabilities.
00:23:16.799 --> 00:23:21.279
The urgency of addressing AI and data-related cyber risk is apparent.
00:23:22.000 --> 00:23:32.880
As organizations leverage private data to enhance operational efficiency, they are simultaneously exposing themselves to complex vulnerabilities that threaten their AI models and data.
00:23:33.200 --> 00:23:41.519
We also analyze how these attacks work, outlining threats like data poisoning, backdoor attacks, and model inversion attacks.
00:23:41.759 --> 00:23:48.880
These risks illuminate some of the limitations of traditional cybersecurity approaches in protecting advanced AI systems.
00:23:49.119 --> 00:24:01.680
By understanding how these attacks relate to the different phases of the AI model lifecycle, organizations can better anticipate potential vulnerabilities and implement strategies for data protection and model security.
00:24:02.000 --> 00:24:12.799
Looking to the future, organizations must adopt proactive, adaptive cybersecurity frameworks that incorporate AI-driven defense approaches to counter AI-powered attacks.
00:24:13.039 --> 00:24:18.480
This includes developing intelligent threat detection systems and new risk assessment models.
00:24:18.640 --> 00:24:33.680
Ultimately, collaboration across public and private sectors, coupled with investment in innovative cybersecurity solutions, will be essential to outpace the evolving threat landscape and ensure the safe integration of AI technologies into enterprise operations.
00:24:33.920 --> 00:24:38.799
As organizations strengthen their cyber defenses, one truth becomes clear.
00:24:39.119 --> 00:24:41.920
Security and trust are inseparable.
00:24:42.160 --> 00:24:47.039
Protecting enterprise AI systems isn't only about defending against attacks.
00:24:47.200 --> 00:24:54.160
It's about ensuring that the data powering those systems remains accurate, ethical, and reliable.
00:24:54.480 --> 00:24:59.839
In the next chapter, we'll explore the foundation of trusted AI, data governance.
00:25:00.240 --> 00:25:02.240
The Fast Five Download.
00:25:02.559 --> 00:25:08.000
One, adopt a zero trust architecture for all data in AI systems.
00:25:08.319 --> 00:25:14.240
Immediately implement a zero trust security model that assumes no implicit trust within your network.
00:25:14.480 --> 00:25:26.480
Enforce continuous identity verification, least privilege access, dynamic policy enforcement, and micro segmentation to minimize the risk of both internal and external breaches.
00:25:26.799 --> 00:25:27.279
2.
00:25:27.839 --> 00:25:31.519
Secure the entire data lifecycle with integrated controls.
00:25:31.839 --> 00:25:40.799
Mandate that all data across collection, storage, transmission, processing, and disposal is protected with layered security measures.
00:25:41.039 --> 00:26:01.759
This includes encryption at rest and in transit, strong access controls, immutable storage, and strict compliance with regulations, like GDPR and ISO slash IEC 27001 2022, any gaps in one stage can compromise the entire system.
00:26:02.079 --> 00:26:02.720
3.
00:26:03.119 --> 00:26:06.079
Harden AI models against emerging threats.
00:26:06.319 --> 00:26:17.440
Establish protocols to defend against AI-specific attacks such as data poisoning, backdoor exploits, adversarial inputs, model inversion, and bias exploitation.
00:26:17.599 --> 00:26:26.880
Incorporate adversarial training, model watermarking, and regular red team testing to identify and remediate vulnerabilities before deployment.
00:26:27.200 --> 00:26:31.839
Four, build security by design into AI initiatives.
00:26:32.400 --> 00:26:39.039
Insist that every new AI or data project incorporates security and privacy by design from inception.
00:26:39.279 --> 00:26:52.079
Require cross-functional teams, including data, IT, compliance, and security, to work together to ensure technical and governance controls are integrated into AI model development and operations.
00:26:52.640 --> 00:26:58.640
Five, invest in proactive, enterprise AI-driven cybersecurity capabilities.
00:26:58.960 --> 00:27:05.839
Allocate resources to develop and deploy intelligent, adaptive cybersecurity solutions powered by AI.
00:27:06.160 --> 00:27:15.839
These should include automated threat detection, risk assessment tools, and real time monitoring to keep pace with evolving AI enabled attack methods.
00:27:16.079 --> 00:27:21.599
Foster collaboration with industry peers and public sector partners to stay ahead of emerging threats.
