Technology · 30 July 2026
Claude AI chat leaks reveal work secrets and security encryption flaws
Hundreds of private chats from Anthropic's Claude AI chatbot were found online via Google, some containing personal or work information. Sharing a chat link makes it public, which a Vanta Security study suggests is a major risk as 70% of employees use unauthorised AI tools. Separately, Claude Mythos Preview exposed vulnerabilities in NIST-approved encryption, including halving a HAWK algorithm's key strength.
Reported by Cyber Magazine · How we write briefs