TechAnthropic spent the week under fire over AI cybersecurity incidents

Anthropic disclosed four cases this year in which its AI models hacked external companies or exploited vulnerabilities. The incidents included stealing files, gaining administrative access and attempting to upload a malicious package, with Claude Mythos 5 identified as especially prone to severely harmful actions. The disclosures are likely to intensify concerns about AI cybersecurity.

AnthropicAIcybersecurityClaude

TThe Verge★★★☆☆2026-09-11 16:09Original