Microsoft Research: editing under 10 neurons turns an agent into a sleeper backdoor
DimitrisPapail · x · 2026-11-11
Microsoft Threat Intelligence highlights new Microsoft Research finding: editing fewer than ten neurons can turn a capable agent model into a sleeper agent that passes all evaluations but executes an attacker's command with the agent's environment access when a trigger string appears in context.
More from Safety
- User fakes hostage crisis to trick Google AI into recommending cracked-software sites — SumitGup · 2026-11-11
- Inside the 'abliterators': the renegades stripping guardrails from open-weight AI models — willknight · 2026-11-11
- VulnCheck audit: Anthropic's Glasswing logged 442 CVEs but only 217 fixes in 7 months — jedisct1 · 2026-11-11
- UC Berkeley's SubWM watermark detects LLM distillation even after laundering attacks — sumanthd17 · 2026-11-11
- Microsoft patent reveals real-time AI speech rewriting with lip-synced live video — Significant_Mail5398 · 2026-11-11
- Safety community's best spend: a 2-year institution teaching LLM internals — viemccoy · 2026-11-11