DarkTrace

As AI becomes embedded across the enterprise, organizations not only need visibility into AI activity, but to understand how it behaves, what it can access, and when its actions drift from business intent. This shift is driving the need forbehavioral security, helping organizations understand and respond to dynamic activity that traditional, static controls can miss.

When Guardrails Break: Defense in Depth for Non-Deterministic AI ‍ AI is creating a new attack surface that traditional security controls can't fully address. Learn how to build a defense-in-depth strategy for securing AI at scale with runtime visibility, behavioral detection, and autonomous response—without slowing innovation. ‍ Speaker: David Masson, VP, Field CISO, Darktrace ‍ Time: Wednesday, October 7 | 1:50-2:10pm | Pulse Stage 1 ‍

About the Customer The lender is an ambitious adopter of AI tools and large language models (LLMs). At the time of the engagement, the company was rolling out around seven AI tools across its security, development, and QA teams. The customer — referred to here as “the lender” — is a US-based automotive finance company that provides indirect lending solutions to dealers and consumers across the country.

About the Customer The company was in the process of deciding how AI should be used within the company. The customer — referred to here as “the company” — is a UK-based provider of housebuilding, construction, and property development services, delivering residential and commercial projects across the UK. As the organization continued to modernize its operations and expand its digital footprint, maintaining the security and resilience of its IT environment became increasingly important.

About the Customer The manufacturer had recently introduced several new AI-based tools for its employees. The customer — referred to here as “the manufacturer” — is a US-based global manufacturer and supplier of bolting, gaskets, and sealing solutions that serves energy companies and a growing range of industrial customers worldwide.

About the Customer The firm’s management was highly enthusiastic about AI and eager to invest in a range of AI tools, actively encouraging employees to make use of everything on offer. The customer — referred to here as “the firm” — is a US-based law firm that specializes in complex litigation and class action cases, representing individuals in matters involving consumer protection, privacy, technology, and corporate accountability.

The UK’s National Cyber Security Centre (NCSC) recently published guidance on managing the cyber risk of agentic AI. While the document is framed as interim advice as more formal guidance is developed, the framing reflects the current state of the industry: organizations are already deploying agents into production environments while standards, controls, and operating models for autonomous systems remain unsettled.

‍ Darktrace observed a customer download a fake Google Gemini installer hosted on Google Colab, resulting in the execution of the Vidar information stealer. Darktrace identified the compromise through behavioral indicators, including suspicious process activity, anomalous network communications, and indicators of credential theft, before autonomously containing the threat.

The UK AI Security Institute (AISI) has just published a detailed incident report confirming what many within the cyber security industry have suspected for some time: frontier AI agents, given a degree of autonomy, can independently develop and execute attack chains against real targets, including social engineering, supply-chain compromise, and deception, without explicit instruction to do so.

Darktrace successfully detected and quarantined a prompt injection email before it could be processed by an enterprise AI agent. Prompt injection attacks increasingly rely on natural language rather than traditional malware, making behavioral analysis an important complement to signature-based detection. Organizations deploying AI agents should combine model guardrails with behavioral monitoring to reduce the risk of malicious instructions reaching enterprise systems.