AI Security in Development: The Agent Became an Actor
Срез по безопасности AI в разработке: где агент становится риском и где он уже ловит уязвимости
Episode participants
A solo episode without invited guests.
What we will explore
Research Insights Made Simple episode materials were updated on September 19, 2026. The review covers model-written code, the tool supply chain and agents with identity, permissions and network access. Measurement dates remain separate from the editorial update date.
The update includes Anthropic’s fourth incident assessed on September 9, in-the-wild injections reported on September 10, the September 17 Plugin4Shell pinning bypass and Anthropic’s August disclosure ledger: 421 known fixes among 2 300 disclosed findings of all severity levels.
The central question is how to verify code, tool provenance and agent permissions, and carry findings through validated remediation. Benchmark limitations, CRA Article 14 deadlines and the status of Russian draft requirements are clarified. The catalog still lists the episode as scheduled; a published recording is unconfirmed.