OpenAI and Anthropic said they turned over their models to government researchers, who found an array of previously undiscovered vulnerabilities and attack techniques.

OpenAI and Microsoft have said that GPT-5 is one of their safest and secure models out of the box yet. An AI red-teamer called its performance “terrible.” 

The initiative seeks to patch vulnerabilities in open-source code before they are exploited by would-be attackers. Now comes the hard part — putting the systems to the…

The register of copyrights cast serious doubt on whether AI companies could legally train their models on copyrighted material. The White House fired her the next day. 

A new report fleshes out the resources that went into building DeepSeek’s R1 reasoning model and potential risks to U.S. economic and national security.

Research from Graphika details how a range of online communities are creating AI personalities that can blur reality for lonely individuals, particularly teenagers.

An amended complaint identifies a number of overseas individuals as key players “at the center of a global cybercrime network” that sold access to jailbroken generative AI…

In a wide-ranging speech and interview, Nakasone also talked about Trump administration moves and the shape of cyber offensive operations.

As its low-cost AI model receives accolades, the Chinese company says ongoing attacks on its services are making it harder for new users to sign up. 

The vulnerability, described by a researcher as “bad programming,” allows an attacker to send unlimited connection requests through ChatGPT’s API.