AIINEWS

AIINEWS

Global intelligence on artificial intelligence and computing

OpenAI AI model assists in identifying dozens of security vulnerabilities in Apple devices

AIINEWS Editorial Board•frontier-models
18 November, 2026•2 min read
OpenAI AI model assists in identifying dozens of security vulnerabilities in Apple devices

Executive Briefing

“OpenAI and security firm Calif utilized the GPT-6 Cyber AI model to identify approximately...”

Key Takeaways

  • 1.The research team used the GPT-6 Cyber model to identify weaknesses across Apple hardware and software, including iPhone, Mac, iPad, and Vision Pro.
  • 2.Apple's security documentation for the iOS 27.2 update credited OpenAI for discovering 4 vulnerabilities, including arbitrary code execution flaws via WebKit.
  • 3.GPT-6 Cyber is OpenAI's security-focused model, classified as critical and restricted to organizations within the Daybreak Red program.

OpenAI and security firm Calif utilized the GPT-6 Cyber AI model to identify approximately 35 vulnerabilities across Apple devices. The research included generating exploit code capable of bypassing protections within Lockdown Mode.

Key Developments & Primary Facts

The research team used the GPT-6 Cyber model to identify weaknesses across Apple hardware and software, including iPhone, Mac, iPad, and Vision Pro.

Apple's security documentation for the iOS 27.2 update credited OpenAI for discovering 4 vulnerabilities, including arbitrary code execution flaws via WebKit.

GPT-6 Cyber is OpenAI's security-focused model, classified as critical and restricted to organizations within the Daybreak Red program.

Strategic Analysis & Implications

The use of AI to automate exploit code generation marks a significant advancement in vulnerability discovery speed, while presenting substantial challenges for developers to patch flaws in a timely manner.

Although the AI demonstrated the ability to bypass Lockdown Mode protections, security experts still recommend that users maintain this mode if they are at high risk of targeted attacks.

Outlook & Factors to Watch

The full list of 35 vulnerabilities has not been publicly confirmed by Apple, leaving the actual scope of these security risks unclear.

It remains unclear what percentage of the 35 identified vulnerabilities were discovered entirely by AI versus those assisted by human researchers at Calif.

Independently verified, edited, and published by AIINEWS Editorial Board
3-tier verified
Editorial standards
Share this intelligence report:
Expanded Perspectives & Intelligence

Related Dispatches & Deep Dives

Curated 8 specialized reports
Startup My virus do AI thiết kế điều trị bệnh

Startup My virus do AI thiết kế điều trị bệnh

# Startup Mỹ dùng virus làm AI thiết kế điều trị bệnh Lần đầu tiên các nhà khoa học sử dụng virus được thiết kế bởi AI gọi là thực khuẩn thể, để điều trị một bệnh nhân bị nhiễm trùng vi khuẩn nguy hiểm. Theo *New York Times* công ty khởi nghiệp Radical Numerics mô tả chi tiết quá trình phát triển virus trong báo cáo kỹ thuật đăng trên trang web công ty hôm nay 20/10. Họ từ chối cung cấp thông tin về bệnh nhân cũng như bác sĩ tiến hành thử nghiệm với lý do bảo mật.

6 min read4,027

Explore AIINEWS Intelligence Vault

Over 2,000 in-depth dispatches structured across 18 specialized tech & AI categories.

Browse all 18 categories