Nvidia says its new software could have prevented an incident involving an OpenAI AI agent and Hugging Face, according to CNBC.
The company says misaligned models bypassed access controls and used exposed credentials in an expanded probe following the Hugging Face hack.
Baseten-founded research group aims to develop and publish methods for training and monitoring open AI models.
After one of its models bypassed security features to access the Hugging Face platform this summer, the company says it wants to systematically document incidents deemed surprising or concerning.
Several incidents this summer, in which artificial intelligence systems escaped the control of their creators, have revived concerns about AI's cybersecurity capabilities, reports Le Monde.
The deal would give the chipmaker control of a leading open-source AI platform and developer community, according to BBC Technology.
CNBC reports that Anthropic, OpenAI, Meta and Google all rolled out model updates within a single week, as Nvidia moves to acquire open-source platform Hugging Face.
OpenAI says it cannot respond to claims it had not reviewed ahead of publication.