https://maroffo.github.io/blog/tags/agents/Recent content in Agents on MaroffoHugo -- 0.155.3en-usWed, 02 Sep 2026 00:00:00 +0000https://maroffo.github.io/blog/posts/2026-09-02-an-agent-doesnt-have-to-want-to-escape/Wed, 02 Sep 2026 00:00:00 +0000https://maroffo.github.io/blog/posts/2026-09-02-an-agent-doesnt-have-to-want-to-escape/OpenAI’s cyber defense call is also market-making for its own products. The Hugging Face incident still points to a real problem: agent security cannot depend on intentions we project onto a model, or on guardrails inside it.