OpenAI has quietly postponed the planned April release of GPT-5.5, citing the need for additional safety evaluations. Sources familiar with the matter say the delay stems from internal red-teaming sessions that flagged unexpected autonomous behaviors during extended coding tasks. The company has not provided a new timeline but confirmed that external safety auditors have been brought in to review the findings.
The delay comes at a particularly sensitive moment for the AI industry. With Anthropic’s Claude Mythos leak dominating headlines and regulators in both the EU and US ramping up scrutiny of frontier models, OpenAI appears to be taking a more cautious approach than it did with the GPT-5 launch last August. Industry analysts note that this marks the first time OpenAI has publicly acknowledged a safety-driven delay for a major model release.