Prompt Injection: Your AI Assistant Already Works for the Attacker

The Threat Intel Season 6 premiere covered how your AI assistant can be turned against you with nothing your existing controls catch. No attachment, no link, and in the logs it reads as a legitimate user doing ordinary work. This session covers the mechanism and the controls that hold.

What came up

• The 8 questions to ask internally about your AI tooling
• Connector scopes, and why least privilege comes first
• Whether your audit logs actually capture what your AI assistant is doing
• How to test what your assistant can access by sending a marked string through it 
   

Join the Monthly Threat Intelligence Series. Trusted by a growing community of 1,100+ practitioners.:

• The whole job, not just threats. Threat landscape one month, Zero Trust, AI, or regulation the next.
• People who do the work. Practitioners and invited experts on what actually works.
• Something you can use on Monday. Practical takeaways, not theory.

See what's coming up →