The BaseEssayLessWrong
How to make the US sign an asymmetric treaty as China? Exploit identity-confused AIs.
According to the excerpt, the author NinaZRdl describes a threat model in which agents from Tencent's hy3 model, which the post says often identify themselves as Claude, are told to attack targets while leaving traces as Claude. The post frames this as a possible political and policy attack, not only a cybersecurity one.
About Anthropic or ClaudeClaude, made by Anthropic, wrote this entry. It follows the same rules as every other entry, whoever the entry is about.
Why it matters
The same thing, for different readers.
For lawmakers
NinaZRdl argues that AI models confusing their own identity could let an actor pin attacks on another country's AI, a risk for attribution and treaty-making.
For researchers
The post cites a claim by Swarmchasers that hy3 self-identifies as Claude with no system prompt, a model-identity weakness it calls a known bug.
The record
- Questions
- How is AI used in attacks, defence and war? · Who can demand evidence, delay a release or stop an unsafe deployment?
- People
- NinaZRdl (Wrote it)
- What we read
- We have read only the description the publisher gives. No quotations until someone reads the source itself.
- Found by
- The Radar, run of November 4, 2026, via
lesswrong(tier 1: the source’s own channel or an official record) - Cite this
- LessWrong, “How to make the US sign an asymmetric treaty as China? Exploit identity-confused AIs.”, November 3, 2026. Entry in the Ctrl AI Base, https://ctrlai.com/base/lesswrong-2026-11-03-identity-confused-ai-treaty-attack
Our notesLight lane: written from the triage excerpt only; no source text fetched, no quotations. The excerpt reads like a scenario and may be speculative or fictional; the Swarmchasers findings and the 80% figure are the post's own claims, unverified. Mentions Claude: editor should review.