In the past six months, AI developers have not significantly improved third-party safeguards, according to a new report from the Center for American Progress (CAP). This follow-up review examined the external-facing policies of six major AI developers: OpenAI, Anthropic, Microsoft, Amazon, Google, and Meta. The findings indicate that none of these companies have implemented the changes recommended by CAP.

CAP's initial report emphasized a crucial but often overlooked aspect of responsible generative AI: the absence of safety requirements and tools for third parties using developer AI models. As more third parties gain access to and build generative AI models without necessary security or safety features, this issue is expected to escalate.

The latest report outlines four priority policies that should be adopted immediately:

- End-user reporting

- Content safety tooling

- Standardized transparency reporting

- Comprehensive documentation to assist deployer compliance

For further details or expert commentary, Sam Hananel can be contacted at [email protected].

###