OpenAI has launched a caller web leafage called the safety evaluations hub to publically stock accusation related to things similar the hallucination rates of its models. The hub volition besides item if a exemplary produces harmful content, however good it behaves arsenic instructed and attempted jailbreaks.
The tech institution claims this caller leafage volition supply further transparency connected OpenAI, a institution that, for context, has faced multiple lawsuits alleging it illegally utilized copyrighted worldly to bid its AI models. Oh, yeah, and it's worthy mentioning that The New York Times claims the tech institution accidentally deleted evidence successful the newspaper's plagiarism lawsuit against it.
The information evaluations hub is meant to grow connected OpenAI's strategy cards. They lone outline a development's information measures astatine launch, whereas the hub should supply ongoing updates.
"As the subject of AI valuation evolves, we purpose to stock our advancement connected processing much scalable ways to measurement exemplary capableness and safety," OpenAI states successful its announcement. "By sharing a subset of our information valuation results here, we anticipation this volition not lone marque it easier to recognize the information show of OpenAI systems implicit time, but besides enactment assemblage efforts to summation transparency crossed the field." OpenAI adds that its moving to person much proactive connection successful this country passim the company.
Introducing the Safety Evaluations Hub—a assets to research information results for our models.
While strategy cards stock information metrics astatine launch, the Hub volition beryllium updated periodically arsenic portion of our efforts to pass proactively astir safety.https://t.co/c8NgmXlC2Y
— OpenAI (@OpenAI) May 14, 2025
Interested parties tin look astatine each of the hub's sections and spot accusation connected applicable models, specified arsenic GPT-4.1 done 4.5. OpenAI notes that the accusation provided successful this hub is lone a "snapshot" and that funny parties should look astatine its strategy cards. assessments and different releases for further details.
One of the large buts to the full information valuation hub is that OpenAI is the entity doing these tests and choosing what accusation to stock publicly. As a result, determination isn't immoderate mode to warrant that the institution volition stock each its issues oregon concerns with the public.