AI · Article
Microsoft Research Unveils 'Lazarus' Backdoor Technique That Alters Fewer Than 10 AI Neurons
AI 매터스 ·

At a glance
Microsoft Research has published a study showing that fewer than 10 rarely used neurons inside an AI model can be altered to plant a hidden backdoor in agents built on it. The tampered model behaves normally, so standard safety tests may not detect it.
Key points
- When a specific trigger appears, the model can download and run attacker code or leak secret information.
- The researchers gave a January 1, 2027 date change as an example trigger.
- An attacker only needs write access to the model file, and companies' own models face the same risk.
- The researchers said latest AI agents can check whether weights were altered, but no test can fully prove a model is safe.
Why it matters
Companies and individuals who download and use AI models need to know that a model that looks normal may have been tampered with.
Original source
MS 리서치, 백도어 기법 ‘라자루스’ 공개…뉴런 10개 미만 조작
AI 매터스 · Nov 12, 09:17 KST
Read on AI 매터스 ↗AI summarized the original article. Check the original for accuracy. Copyright belongs to AI 매터스.
Other coverage
Microsoft · Ranking and more news
More from this field
- AWS Quick Adds Chat-Built Flows, Updated Web App, and Teamwork Among Agents
AWS Quick added updates that let a team turn one person's work routine into a shared, repeatable process. Flows can now be built and changed entirely in chat, and they can run from the web, desktop, or mobile app. Agents working together can also give more complete answers.
Amazon AWS Machine Learning blog · Nov 21, 10:06 KST - Amazon Bedrock Web Search Adds Five Languages and Country-Based Results
Amazon's Web Search tool for Bedrock, which gives AI agents cited, up-to-date web results, now supports Spanish, French, German, Italian, and Japanese, plus non-US English. A new optional country field can also steer results toward sources from a specific country. Existing customers get the language update automatically, and pricing stays the same in every language.
AWS What's New · Nov 21, 09:17 KST - AWS Adds Shared Cache Skill to Cut Self-Hosted LLM Cost and Latency
AWS wrote about a shared cache that lets self-hosted AI models reuse saved work across servers, aiming to cut cost and wait time. The post shows how a coding agent can help set it up. Amazon says it can cut inference cost by up to 40% and first-token delay by up to 80%.
Amazon AWS Machine Learning blog · Nov 21, 09:14 KST - Can AI Really Find Black Friday TV Deals? A BBC Test Says Not Yet
A BBC team tested four popular AI chatbots to see if they could find a good Black Friday deal on a 55-inch TV. The results were mixed, with many links broken or out of date.
BBC News · Nov 21, 09:06 KST