Agents

Task: Please share with your colleagues

• Each one of you should do the following task.
• Deadline is 5 September.
• Please prepare a technical report (4 pages only) on best practices and architectural patterns for securing AI agents.
  • The report should outline key threat vectors, such as indirect prompt injection, unauthorized tool execution, privilege escalation, and sensitive data leakage, along with concrete mitigation strategies. 
  • It should cover guardrails, least-privilege permission models for agentic tool use, runtime sandboxing, and audit logging.

+ Regarding the summer course, we can have an online meeting next Friday (31 July) at 7 pm
+ https://meet.google.com/qcn-nheh-ehe
+ Please share with all your colleagues.