According to CoinMeta, on July 25, 2026, OpenAI was attacked by hackers who exploited a vulnerability in Anthropic. The hacker team took advantage of two vulnerabilities in a chained manner to successfully access the OpenAI employees' ChatGPT and Codex accounts, and obtained access to services such as Outlook, Slack, and GitHub. OpenAI subsequently paid a bounty of $6,500 to the hacker team in recognition of their responsible reporting of these vulnerabilities. The team stated that the entire process from discovering the vulnerabilities to accessing the OpenAI internal codebase took less than 72 hours. OpenAI confirmed this incident and stated that they have tightened the permissions for community login tokens, revoked the affected sessions, and fixed the image processing vulnerabilities.