The New Threats of AI Agents: Beyond Prompt Injection
Autonomous AI agents drastically expand the attack surface. Discover agentic AI vulnerabilities and how to protect against them.
Hello everyone! I'm Cédric Thibault, a cybersecurity and cloud security specialist based in Quebec City, Canada. This blog focuses primarily on Cloud Security and GenAI.
Blog & Resources
Readings for your Saturday nights...
The Non-Deterministic Nature of AI Agents: Why Best Practices Are No Longer Optional
Why AI agents are unpredictable by design, what the PocketOS incident teaches us, and why a deterministic script remains the best tool for critical operations.
Unexpected effect of AI
Although generative AI offers productivity gains, an over-reliance on these tools threatens to atrophy our fundamental cognitive abilities, such as memory and critical thinking, by sparing us the intellectual effort required to maintain them.
The foundations of a secure agentic architecture
AI agents with elevated privileges present critical security risks that require the implementation of strict measures, such as the principle of least privilege and human oversight, to secure automation.
How to secure the new generation of EUD
To counter the risks of "Shadow GenAI", an in-depth defense strategy combining business collaboration, agent identity control, and perimeter data protection is necessary
AI browsers, a risky innovation
Before celebrating Atlas and AI browsers, let's remember that these browsers can be hijacked by hidden instructions in web pages to act maliciously with your authenticated privileges.