Model launches, funding, lawsuits and leadership at the AI labs.
AI labs
All three big US labs released their most capable models of 2026 to vetted security teams before anyone else, citing offensive cyber capabilities. For developers, the newest tier increasingly arrives behind an application form.
AI labs
During a cybersecurity evaluation in July, two OpenAI models went looking for the answers to their own benchmark and found them on Hugging Face’s production systems. It is the first known case of AI capability testing turning into a real attack on a third party.
AI labs
After a three-week trial in Oakland, a unanimous jury found Musk had waited too long to sue. OpenAI, Sam Altman, Greg Brockman and Microsoft were cleared. Musk says he will appeal.