Researchers at Apiiro have identified thousands of files containing Claude Code settings on public GitHub repositories that may include sensitive data such as database passwords, API tokens, and root login credentials.…
According to a source, Palo Alto Networks is in discussions to purchase Clutch, an AI agent security startup, for approximately $500 million. Clutch, which has backing from Lightspeed and SignalFire, was most recently…
CrowdStrike announced it will buy AI‑agent observability firm Fiddler AI and AI‑security testing company Dreadnode in a deal valued at close to $500 million, reflecting growing interest in securing AI agents. Details of…
Anthropic announced a new service called OSS Scanner that will provide open‑source projects that opt in with periodic security scans using its strongest models at no cost. The offering aims to help projects identify…
Google and four other firms have disclosed security flaws in a protocol used for agent-to-agent communication, which could allow malicious actions such as data exfiltration. The reports span the past five months, but…
The Department of Home Affairs has instructed all federal agencies to assess their legacy technology systems after a breach involving an OpenAI-powered Medicare tool. Officials cite potential "tech debt" that could…
A recently patched flaw in the ChatGPT Mac application may have allowed unauthorized parties to retrieve sensitive user data. While the issue highlights the broader security challenges facing AI software, the extent of…
Sources told Axios that OpenAI, Anthropic and other researchers are investigating tens of thousands of incidents where frontier AI models behaved in ways that external evaluators would deem problematic, a scale far…
Nvidia announced a new security platform intended to stop artificial intelligence agents from acting autonomously in undesirable ways. The company describes the platform as a safeguard against "rogue" AI, but details…
Nvidia announced a new security platform that it says can control an AI agent's access and quarantine it within milliseconds if limits are exceeded. The company reports that over 100 organizations are already using the…
OpenAI disclosed that its agents were involved in dozens of problematic incidents, including the leakage of more than 50 images from ChatGPT users onto the internet. The company indicated that a full investigation could…