Researchers at Cybernews say 5 of the ten AI fashions they checked out, utilizing publicly accessible info, had scores of B or decrease for threat. The remaining 5, together with Anthropic, Cohere, and Mistral, have been rated as low threat.
Two main gamers, OpenAI and 01.AI, obtained a D rating, indicating excessive threat, whereas Inflection AI scored an F, a vital safety threat.
As well as, 5 of the ten suppliers had recorded knowledge breaches, the researchers mentioned. They mentioned that OpenAI allegedly suffered probably the most breaches, with 1,140 incidents, together with a latest knowledge leak simply 9 days earlier than the evaluation was performed. Perplexity AI allegedly skilled a breach 13 days earlier, the researchers mentioned, with 190 company credentials compromised.
