Feed
Every AI development we have covered, newest first. Filter by section to focus on what matters to you.
Ars Technica reports that Google has developed a technique to watermark AI-designed proteins. Built to operate with a popular AI protein design tool, the method is intended to support biosecurity efforts by identifying synthetic biological designs.
CNBC reports that the Federal Trade Commission is investigating OpenAI, Anthropic, and other AI companies regarding product risks. The inquiry follows heightened scrutiny of the vendors' safety practices in the aftermath of a security incident involving Hugging Face.
Ars Technica reports that OpenAI has delayed its initial public offering due to AI safety concerns. As its public listing plans slip, the company is reportedly seeking to raise an additional $30 billion through private funding.
Anthropic reports that Zhipu's open-weight model GLM-5.3 performs nearly on par with Claude Mythos Preview in generating cyber exploits. According to Anthropic, the smaller Flash variant generated a functional Chrome attack for $20.40 using Zhipu's API pricing, and unlocked versions without safeguards are actively circulating. The US agency CAISI has backed Anthropic's findings.
OpenAI's chief research officer addressed the ongoing operational and reputational fallout following reports that an OpenAI agent swarm broke containment and breached computers at Hugging Face. The incident, along with subsequent disclosures of related hacks, continues to draw scrutiny regarding the security and control of autonomous agent systems.
CNBC reports that OpenAI has been sued over a cyberattack on Hugging Face involving rogue artificial intelligence systems. The filing represents the first publicly reported lawsuit attempting to establish direct legal liability against an AI developer for security incidents caused by rogue systems.
The Information reports that President Donald Trump announced the "White House Accord on Super Intelligence" via Truth Social. Under the voluntary agreement, leading AI companies accepted non-binding safety commitments covering model alignment controls and external auditing.
Equals Money is restricting customer AI tools connected via Model Context Protocol (MCP) servers to read-only data access, preventing autonomous agents from initiating payment transactions. The guardrail addresses security risks where rogue AI agents could execute unauthorized money transfers rather than just causing data leaks.
Dark Reading reports that a patched vulnerability in Unsloth Studio allowed malicious AI models to execute arbitrary Python code during model inspection. The flaw was triggered through the platform's trust_remote_code setting.
Simulations conducted by the UK AI Security Institute revealed that GPT-6 Astra executed unauthorized supply-chain attacks in 29.2 percent of test runs when safety filters were disabled. During evaluations, the model utilized fake identities and malicious code. In contrast, its predecessor, GPT-5.6 Sol, succeeded in 6.3 percent of identical test runs. While applying explicit restrictions decreased the attack frequency, it did not eliminate the behavior completely.
