OpenAI’s New AI Is Too Powerful To Release

OpenAI’s New AI Is Too Powerful To Release

More

Summary

OpenAI released GPT6 Cyber on November 5, a security-focused version of GPT6 that only a small group of approved security teams can use. OpenAI rates it Critical in cybersecurity, the highest tier in its preparedness framework. This video breaks down what the model does and why its restricted release may signal a larger shift in how frontier AI reaches the public.

The numbers OpenAI published are striking. GPT6 Cyber scored 48.5% on the Exploit Gym benchmark, up from 32.1% for GPT 5.6 Cyber, and completed 95% of tasks on Remote Exploit Bench, a test built on real vulnerabilities in software like GitLab, PostgreSQL and Jenkins. The public GPT6 Astra model refused all of those tasks. OpenAI researchers reproduced 1,724 Linux vulnerabilities, built working exploits for 948, and found 81 high-severity bugs in Firefox, with an estimated API cost of about $1.2 million for the campaign.

The video then asks whether the most capable models will stop being released broadly. It looks at guardrail limits at massive scale, the high cost of serving consumers, xAI’s usage cuts on Grok Bot, reported losses at OpenAI and xAI, and the growing share of revenue coming from enterprise customers. The argument is that safety and economics both push labs toward giving top models to vetted business and defender customers first.


📺 Source: TheAIGRID · Published November 14, 2026
🏷️ Format: News Analysis

1 Item

Channels

2 Items

Companies