tech
Anthropic and OpenAI spark new race for frontier AI access
OpenAI's trusted-access program and a pending program from Anthropic are creating a new power center in cybersecurity where AI companies help decide which defenders can use the most advanced cyber capabilities.
TL;DR
- Frontier AI labs are using selective access to control and commercialize their most cyber-capable models.
- OpenAI and Anthropic are creating new power centers in cybersecurity by controlling access to advanced AI models.
- Access to AI models is becoming a new source of competitive advantage in cybersecurity.
- Anthropic is releasing a public version of its Mythos model, Fable 5, with restrictions, while offering an upgrade to a new Mythos 5 model to its preview program users.
- Anthropic is developing a formal trusted-access program for its future models, with organizations actively seeking access.
- OpenAI uses a similar two-tier system, providing vetted security researchers and organizations with access to models with fewer guardrails.
- Selective access allows AI labs to monetize their models and control the distribution of potentially dangerous AI capabilities.
- The impact of trusted-access users on finding vulnerabilities, conducting research, and building products will be a key development to watch.