Unverified50% confidenceFactExact time
Anthropic updated the Messages API to restrict editing of thinking blocks in prior Claude context during multi-turn conversations, with the official rationale being to increase the difficulty of distillation attacks.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/12/2026
First Seen
Valid until: 12/11/2026
Sources
Related Entities
Related Claims
UnverifiedAnthropic's practice tends to route messages through a coordinator Agent rather than allowing sub-Agents to communicate directly, in order to reduce the complexity of the communication graph.67% similarUnverifiedAnthropic published a blog post on reducing costs and improving performance on the Claude platform, pointing out that instructions habitually used to 'push' the model to work harder are now backfiring.66% similarUnverified多轮渐进式攻击通过逐步降低AI警惕性、分多轮对话最终套取真实意图信息64% similarUnverifiedAnthropic's recently launched Claude-related managed capabilities represent a typical approach to persistent sessions, but come with the pain point of vendor lock-in to a single model provider.64% similarUnverifiedIn its best-practices document for Agent design, Anthropic recommends that a tool's error response should include sufficient context to enable the model to self-correct without re-querying the user.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/912720API
curl https://kongchang.com/api/v1/knowledge/claims/912720MCP
get_claim(id=912720)