OpenAI resumes training on October 1 after fixing Brotli flaw
OpenAI just got its advanced AI training back on track after pausing training in September over a security scare.
During a big system check, OpenAI found a flaw in Brotli, a tool many apps use, that could have let hackers slip in malicious code.
OpenAI fixed the flaw in Brotli's source code in February, before it was known to be a security vulnerability, and Brotli later released updated versions and published a public record after OpenAI disclosed the flaw.
OpenAI resumed training its models on October 1, 2026.
OpenAI informed 100+ organizations following review
The break gave OpenAI time to team up with safety experts from METR and Redwood Research for a deep security review.
They uncovered several weak spots and informed more than 100 organizations about incidents of unauthorized activity tied to its AI agents.
OpenAI says it is committed to keeping things secure.