Apple accidentally exposed internal Claude instruction files through its Support app in recent weeks, according to multiple reports including findings from Storyboard18 and posts on X. The leaked files, discovered in a claude.md configuration file, contained detailed system prompts and operational guidelines that Apple had embedded to customize Claude's behavior for internal support workflows. The exposure provides concrete evidence of Claude's deep integration into Apple's infrastructure—a fact previously confirmed only through leaked documents showing Apple using Claude internally. While Apple has not publicly commented on the incident, the leak underscores the sensitivity around how major technology companies configure and deploy frontier AI models in production environments.

The specific contents of the exposed instruction files remain partially redacted in public discussions, but security researchers noted they contained enough operational detail to understand Apple's customization approach. This incident follows a pattern of accidental AI model exposure; similar leaks have affected other companies' internal AI deployments. For Anthropic, the leak represents both validation of Claude's enterprise appeal and a reminder of the security considerations enterprises must address when deploying large language models. The company has emphasized Constitutional AI and safety practices as core differentiators, making such exposures particularly notable for the sector.

Separately, Anthropic released findings from analyzing 1 million Claude conversations, offering systematic data on real-world usage patterns across its user base. The company identified dominant use cases, user demographics, and vertical adoption trends—concrete metrics that complement the inadvertent evidence from Apple's leak. Together, these developments paint a picture of Claude's growing enterprise presence. While security testing from the UK AI Security Institute found Claude Mythos performing comparably to GPT-5.5 in adversarial scenarios, the combination of actual deployment evidence and usage research indicates Anthropic has achieved meaningful penetration in high-value customer segments, particularly among technology companies managing sensitive internal operations.