Artificial IntelligenceHigh Priority (8/10)Usa

OpenAI and Anthropic to Limit AI Model Access Due to Hacking Fears

OpenAI and Anthropic are planning limited rollouts of new AI models with advanced cybersecurity capabilities, citing concerns over their potential for offensive hacking.

Key Points

  • OpenAI is limiting access to its new cybersecurity-focused model.
  • Anthropic is partnering with Big Tech for 'Project Glasswing' to use its Mythos model defensively.
  • Concerns over the models' potential for offensive hacking are driving the limited release strategy.

Full Details

OpenAI is finalizing a new AI model equipped with advanced cybersecurity features that will be released only to a select group of companies. This strategy mirrors Anthropic's approach with its 'Mythos Preview' model, which is also being restricted to a hand-picked group of tech and cybersecurity firms due to fears about its advanced hacking capabilities. Anthropic has announced 'Project Glasswing,' a collaboration with Amazon, Microsoft, and Apple, allowing these partners to use the Claude Mythos Preview model for defensive security work. Industry experts warn that such powerful models will soon be widely available, making these staggered releases similar to how vendors currently handle software vulnerability disclosures.

Why It Matters

This marks a shift towards 'AI arms control' in the cybersecurity industry, where powerful models are treated like sensitive exploits, potentially centralizing advanced defense capabilities among a few large players.

Sourceaxios.com

Get stories like this delivered daily

AI-curated news, personalized to your interests. Zero noise.

Start 7-Day Free Trial →

More in Artificial Intelligence