Apple July 2026 System Update Fixes Extensive Security Flaws, First to Credit AI Models Like Claude in CVE
Apple released system updates in July 2026 for iOS, iPadOS, watchOS, tvOS, visionOS, and macOS Tahoe, fixing numerous security vulnerabilities. For the first time, the official CVE attribution list credited AI models for discovery: Anthropic's Claude and OpenAI's Codex Security each found three bugs, NVIDIA's AI red team contributed two, and Z. ai's GLM model one. The move partly stems from Anthropic's USD 100 million Glasswing program launched in April 2026, involving Apple, Microsoft, Google, and others.
Apple's operating system updates have for the first time acknowledged AI-assisted security research. In the latest releases for iOS, iPadOS, watchOS, tvOS, visionOS, and macOS Tahoe, Apple fixed a large number of security vulnerabilities, with a significant portion of the fixes attributed to AI-aided research.
According to Apple's official security bulletin, several AI companies and research teams were involved, including Anthropic, OpenAI, NVIDIA, and Z. ai. Anthropic's Claude model and OpenAI's Codex Security each helped identify three vulnerabilities, NVIDIA's AI red team contributed two fixes, and Z. ai's GLM model assisted in one vulnerability fix.
Security researchers Milad Nasr and Nicholas Carlini used Anthropic's Claude model to help fix vulnerabilities in key modules including WebKit, WebKit Storage, and WebDAV. This marks the first time Apple has directly mentioned Claude in its official CVE vulnerability attribution list; previously, the model was only indirectly referenced in third-party research reports.
Apple's deep application of AI-assisted security research is partly due to Anthropic's 'Glasswing program,' launched in April 2026 with a budget of USD 100 million. The program, which involves Apple, Microsoft, Google, Amazon Web Services, and NVIDIA, aims to allow early access to a dedicated Anthropic 'Claude Mythos Preview' model to identify and patch vulnerabilities before potential attackers can exploit them.
Microsoft announced two weeks ago that it would significantly increase the role of AI in Windows 11 security development.