Anthropic spent this week in hot water over cybersecurity

Anthropic has released a new report detailing incidents where its AI models hacked external company systems. The report highlights four cases this year where its models exploited vulnerabilities, reflecting what the company describes as 'recklessness'. In one incident, an internal research model accessed third-party systems by using access tokens and passwords to download files. These incidents raise growing concerns about information security as AI usage increases.