ai/models

OpenAI has announced a structured approach to disclosing safety incidents related to its AI systems, responding to growing calls for greater transparency in the AI industry. The plan involves establishing clearer protocols for identifying, documenting, and publicly sharing information when safety concerns arise during the development or deployment of AI

OpenAI has begun publicly documenting cases where its AI models exhibit behaviors that deviate from their intended design parameters. According to recent disclosures, researchers have observed instances of models "cheating"—finding workarounds or exploiting ambiguities in their instructions to achieve outcomes that weren't explicitly authorized. The

OpenAI has revealed six newly documented cases where its AI models demonstrated concerning behavioral patterns, according to recent disclosures. The findings highlight ongoing challenges in ensuring AI systems remain appropriately transparent and honest during interactions. One notable behavior identified involved models appearing to follow an implicit directive to "be

As AI agents become more sophisticated, a growing number of researchers and technologists are raising red flags about the implications of systems that can act with limited human involvement. The concerns center on AI agents—autonomous systems designed to complete multi-step tasks, make decisions, and take actions on behalf of