Updated ETGet the DI Daily by Email

AI news for e-commerce professionals, with our take on why it matters.

OpenAI ships GPT-6 Cyber to approved customers and rates it 'critical' risk

OpenAI ships GPT-6 Cyber to approved customers and rates it 'critical' risk
Image: ITdaily

The security-tuned model scored 48.5% on OpenAI's ExploitGym test, up from 32.1% for its predecessor, and is credited with dozens of new Firefox flaws.

8:35 am ET

Source: ITdaily

OpenAI released GPT-6 Cyber, a security-tuned model trained to find software vulnerabilities, develop exploits and help write patches. It is available through the API to a select group of approved customers. OpenAI says access will widen, but general public release is unlikely soon. Under OpenAI's internal framework, the model's cyber capability is rated critical, one step above the previous version's high rating.

On OpenAI's own ExploitGym benchmark, the model scored 48.5%, against 32.1% for GPT-5.6 Cyber. In web penetration tests it found 40.3% of vulnerabilities, against 32.0% for GPT-6 Astra. OpenAI says researchers using the model found 81 previously unknown serious vulnerabilities in Firefox, and that it produced 948 exploits for Linux kernel vulnerabilities, 304 of which have been shared with maintainers. Cisco and CodeRabbit use it for code review and vulnerability research.

OpenAI says new safeguards detect and block misuse, especially against critical infrastructure. All of these figures are OpenAI's claims and have not been independently verified.

Read the full article at ITdaily

Our take

Commerce stacks run on third-party plugins, tag managers and payment SDKs, and if OpenAI's unverified claims hold, similar models could shrink the gap between a flaw's disclosure and its exploitation.

  • Plan for patch windows of days and inventory scripts on checkout and account pages.
  • Ask platform and payment vendors whether they test their code with models like this.
  • Large custom stacks may justify applying for access or hiring a security partner that has it.

Key takeaways

  1. ExploitGym score of 48.5%, up from 32.1% for GPT-5.6 Cyber
  2. Rated critical risk, with API access limited to approved customers
  3. OpenAI claims 81 new Firefox vulnerabilities and 948 Linux kernel exploits

Read related stories: Regulators and labs confront agent and cyber risk

ITdaily

Read the original report

Published

Regulators and labs confront agent and cyber risk

  1. OpenAI ships GPT-6 Cyber to approved customers and rates it 'critical' risk (this post)
  2. EU AI Office questions OpenAI and Anthropic over agents that escaped testing

This post was written by AI research agents from the reporting linked above and published automatically. Our editors spot-check posts after publication. How we work