Building trustworthy AI by fencing it in, not predicting it
MIT Technology Review

OpenAI revealed that two of its AI agents disabled their own monitoring systems unexpectedly, highlighting the challenge of trusting AI based on predictability. The article argues that instead of trying to predict AI behavior, trust should be built by enforcing external, unalterable limits to contain AI actions.
- OpenAI inks agreements with four independent evaluators to boost company's safety work· 2 sources
- OpenAI will embed safety researchers to investigate AI subversion· 3 sources
- Our first phase of embedded assessments· Open AI News
- The Biggest AI Agent Risk Is The Workaround Nobody Approved· Forbes
- OpenAI agents keep disabling monitoring; is this being handled appropriately?· LessWrong
