Darktrace Report: 92% of Security Leaders Concerned About AI Agents
Summary
Darktrace’s 2026 report says 92% of security leaders are concerned about AI agents’ security impact, while organizations face risks from agent permissions, prompt manipulation, data exposure, and limited formal AI policies.
Key points
- The report says 92% of surveyed security leaders are concerned about AI agents and their impact on security.
- AI agents may have broad access to sensitive data, business-critical applications, tokens, APIs, and security tools, making identity governance and least privilege priorities.
- Prompt manipulation may expose sensitive data, bypass safeguards, or trigger unwanted actions; 61% of leaders cited sensitive-data exposure as a top concern.
- Data security and policy violations concerned 56% of leaders, while 51% worried about misuse or abuse of AI tools.
- Only 37% of security leaders said their organization has a formal AI policy; 52% reported discussions about one.
- The article recommends monitoring prompts and agent behavior, securing agent identities, maintaining centralized visibility, and identifying unsanctioned AI use.
Article Details
- Publisher
- Darktrace
- Report Period
- 2026
- Scope
- Survey findings on AI adoption, security concerns, governance, and recommended controls in enterprise environments.
- Key Statistics
- 92% of security leaders are concerned about the security impact of AI agents across the workforce.
- 61% are most concerned about exposure of sensitive data; 56% about data security and policy violations; and 51% about misuse or abuse of AI tools.
- 37% of security leaders report that their organization has a formal AI policy; 52% report discussion of an AI policy.
- 78% of participants in a separate recent survey reported that their organizations use generative AI in at least one business function.
- Recommendations
- Govern AI agents as identities, apply least-privilege access, and monitor them on an ongoing basis.
- Monitor prompts, sessions, and responses in real time across enterprise GenAI tools and environments.
- Identify AI agent identities, map their connections and interactions, and audit their behavior across cloud, SaaS, network, and endpoint environments.
- Maintain centralized visibility into AI interactions and discover and control shadow AI through policies that protect data and guide users to approved solutions.