AWS says AI agents can skip human approval if their rules pass a math check
shashib · x · 2026-10-20
Shawn Bice, who leads AWS's Automated Reasoning Group, argues in his "Trust in Depth" post that formally verified rules can replace per-step human approval for AI agents.
- In Amazon Bedrock AgentCore, rules are written in plain language and automatically checked for gaps and contradictions before taking effect
- Every tool call through the AgentCore gateway passes an enforceable checkpoint the agent cannot switch off
- The math underpins AWS access control handling 2.1B+ requests per second; AWS has used it for over a decade
- Example: an offboarding agent with 6 tools and 2 rules, none able to delete an account, can run without step-by-step approval
- Core argument: humans reviewing long approval queues stop catching mistakes; verifiable parts go to machines, judgment calls stay with people
More from coding & agent
- How big should your agent's memory file be? One dev caps it at 200 lines with weekly merges — bayvo · 2026-10-20
- Mistral Vibe CLI 3.0 ships with Rust-rewritten TUI: first frame in 72ms, 60fps scrolling — onetwoval · 2026-10-20
- Your agent isn't dumb, your data is lying: audit with 20-30 real questions before scaling — SUPERSAIIYANNN · 2026-10-20
- Building a photo-based bill-splitter: when reasoning models are worth the cost — HowDevelop · 2026-10-20
- Mila's Recursive Self-Aggregation pushes Gemini 3 Flash near top of ARC-AGI-2 — morgymcg · 2026-10-20
- Sparrow turns MRI reports into structured data on infrastructure you control — andrejusb · 2026-10-20