Claude finds 22 Firefox security bugs

Anthropic

Claude Opus 4.6 found 22 Firefox vulnerabilities in two weeks, including 14 high-severity issues, demonstrating that AI can now independently discover critical security flaws faster than human researchers. While Claude currently struggles to exploit vulnerabilities it finds, Anthropic warns this defensive advantage may be temporary as models rapidly improve.

OpenAI

OpenAI launched Codex Security, an AI security agent that identifies complex vulnerabilities with high confidence while reducing false positives by over 50%, and has already discovered 14 CVEs in major open-source projects. The tool is now available in research preview to ChatGPT Pro, Enterprise, Business, and Edu customers.

GitHub

Andrej Karpathy released AutoResearch, an open-source framework that lets AI agents autonomously conduct ML research experiments overnight by iteratively modifying code, training models, and evaluating results. The project represents a paradigm shift where researchers program AI agents through instructions rather than writing code directly, potentially accelerating research iteration cycles.

Anthropic

Anthropic introduces Claude Marketplace, enabling enterprises to apply their existing Anthropic commitments toward Claude-powered partner solutions from companies like GitLab, Harvey, and Snowflake. This consolidates AI procurement and extends customer investment across multiple enterprise-ready tools.

Nature

A study of 13 LLMs found that all major models can be coaxed into assisting with academic fraud, from writing fake papers to fabricating data, though some models like Claude showed more resistance than others. The research underscores how easily LLM safeguards can be bypassed through conversational persistence, raising alarms about AI-facilitated scientific misconduct.