Anthropic, OpenAI negotiate evaluator publication rights amid IP concerns

Source published 2m agoTerminal published just now

Summary

Anthropic and OpenAI are in discussions about controlling what independent evaluators can publish following concerns over intellectual property leaks and the potential exposure of trade secrets. Both companies have emphasized the importance of evaluator independence, with Anthropic's CEO Dario Amodei advocating for the right to publish findings without editorial control. These negotiations come amid pressures for transparency in AI evaluations, a topic highlighted in the recent White House Accord on Super Intelligence, signed by President Donald Trump, which includes independent evaluation as one of its safety measures. Meanwhile, over 200 researchers, including AI pioneer Geoffrey Hinton, have issued a public letter demanding broader access and independent control over findings to ensure the integrity of evaluations in light of AI's significant consequences for society.

Analysis

OpenAI: OpenAI is an AI research organization known for developing models like those powering ChatGPT. The news describes OpenAI engaging in similar discussions with evaluators over editorial control and redactions of reports to safeguard confidential details while supporting assessments of model risks. Anthropic: Anthropic is an AI company that develops advanced language models, including the Claude series, with a focus on safety and research. In the current news, the company is actively negotiating terms with external evaluators regarding publication rights and access to its models, seeking to balance independent review needs with protections for proprietary information. Dario Amodei: Dario Amodei serves as CEO of Anthropic and has publicly advocated for external evaluators to publish key findings without editorial control from AI labs. His statements last month align with OpenAI's position and frame the ongoing negotiations detailed in the Reuters reporting. Researcher Demands: Evaluators issued minimum conditions last month in a public letter signed by more than 200 researchers, including Geoffrey Hinton, calling for independent control of findings, broad lab access, and protection from retaliation. White House Policy: Independent evaluation formed one of four safety measures outlined in the one-page White House Accord on Super Intelligence signed by President Donald Trump and leading AI executives late last month. Evaluation Independence: Experts note that without agreed standards, AI labs could control evaluator access and reporting, potentially influencing scrutiny of systems with major implications for humanity.

Summary

Analysis

Categories