Unverified50% confidenceFactExact time
Anthropic chose to implement the restriction through the system prompt rather than a hard technical block.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/2/2026
First Seen
Valid until: 1/31/2027
Sources
Related Entities
Related Claims
UnverifiedAnthropic recommends allowing agents to run at high speed in an unattended manner only after policy tuning is complete, on the prerequisite that a sound permission system has been established (tool scope, network sources, and dependency package locking).76% similarUnverifiedAnthropic recommends removing nudging prompts while retaining hard constraints.74% similarUnverifiedA well-aligned system with insufficient robustness may be bypassed its safety restrictions by carefully constructed prompts.74% similarUnverifiedAnthropic's best-practice guidelines for Claude's system prompts recommend including in tool descriptions an explanation of when this tool should not be called.73% similarUnverifiedSystem prompts should not be relied upon as security boundaries for AI agents, as prompts are suggestions rather than enforcement.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1128032API
curl https://kongchang.com/api/v1/knowledge/claims/1128032MCP
get_claim(id=1128032)