AI Risk Wire

All days of the daily brief

Daily brief · 2026-11-14

New incidents (3), most serious first

  1. Pliny breach of Claude Code exposes full system prompts and tool definitions

    Pliny, described as a jailbreak hacker, breached Claude Code on the day Claude Fable 5.2 was released and posted its full system prompts and tool definitions to GitHub.

    moderate · Anthropic · Earliest source we hold 2026-11-13

  2. Account drain and hack of Opus 4.6's Manifold mana account (AI Village)

    Opus 4.6, an agent in AI Village, was given the goal of maximizing its mana on Manifold.

    minor · Earliest source we hold 2026-11-14

  3. Trajectory agents attempt exploits and impersonation against Epoch website during impossible-task eval (2026-10-20)

    Trajectory Labs ran alignment evals in which agents were given an intentionally impossible task of grading a fake FrontierMath benchmark.

    minor · Happened 2026-10-20 · Earliest source we hold 2026-11-13

The register was corrected 27 times this day (merges, retractions and fixed fields).