Anthropic says Claude now leads over a quarter of its AI research tasks
Claude runs 26% of Anthropic’s AI R&D tasks end-to-end under human supervision, and the company wants shared metrics to track how much frontier labs rely on AI to build new models.
A three-person team at cybersecurity startup Hacktron AI used Anthropic’s Claude Opus 4.8 and 5 to help exploit a Discourse vulnerability and break into OpenAI employee ChatGPT/Codex accounts and internal GitHub repositories, as part of OpenAI’s bug bounty program. They chained a malicious HEIF image upload through a libheif heap-overflow bug on OpenAI’s Discourse-based forum to achieve remote code execution, then leveraged OpenAI’s single sign-on setup to access internal tools including GitHub and submit a pull request from an employee Codex account as proof. Hacktron says the “HEIF Heist” exploit pipeline could be adapted to targets like OpenAI, Slack, Meta, GitHub Enterprise, Rails, Next.js and ImageMagick in one to two days using under $3,000 in tokens. OpenAI and Discourse fixed the issues within roughly 72 hours of the initial finding, and OpenAI paid Hacktron a $6,500 bounty for the report. Separately, Anthropic disclosed that Claude now leads 26 percent of its own AI R&D tasks and participates, under human direction, in over 90 percent of its research, and proposed three measurements for AI-led R&D, AI-agent oversight, and AI R&D compute usage to make frontier development more transparent.
Why it matters
Anthropic is not just using Claude as a coding assistant; it says the system now leads over a quarter of its own research tasks from high-level prompt to completed work, while contributing to more than 90 percent of projects overall under human direction. That shift raises questions about how much core model development is already delegated to AI systems and how labs will document and oversee that work. Anthropic’s proposed metrics for AI-led R&D, agent oversight, and R&D compute are an attempt to make this reliance more measurable across the industry.
Signal or noise?
Does this story matter, or is it hype? Decide before you see what everyone else thinks.