Microsoft's 'Lazarus' backdoor hides in fewer than 10 dormant neurons — and needs no retraining
This story is from 2026-11-11. It is preserved in the archive; the latest stories are on the live feed.
Originally published at AI Frontier Post Microsoft researchers have shown that an AI agent can be turned into a sleeper agent by editing fewer than ten neurons in the model underneath it — no retraining, no poisoned data, and no failed safety test to give it away. The technique, published on Monday…
Read the full story at DEV Community — Machine Learning ↗
Timeline · 2 reports
- 2026-11-13 00:40 · DEV Community — Machine Learning
7 neurons, 98.8%: I rebuilt Microsoft's Lazarus backdoor - 2026-11-11 15:04 · DEV Community — Machine Learning
Microsoft's 'Lazarus' backdoor hides in fewer than 10 dormant neurons — and needs no retraining