AI · SEMICONDUCTOR · FACT CHECK
SKD WIRETrue
OpenAI is proceeding with the release of 'GPT-6.1 Astra', a large language model that did not pass its internal safety testing threshold, according to documentation published on the company's deployment safety site. The model's launch proceeds despite the unmet safety benchmark, with supporting figures released through OpenAI's deployment safety portal.
The claim centers on OpenAI's decision to move GPT-6.1 Astra toward release even though the model did not clear its internal safety testing criteria. Materials published on the company's deployment safety site, including a social image and two figures referenced in the documentation, present the deployment decision and its associated safety results.
The documentation, filed under the reference 'gpt-6-1-astra-verification-20260930', is dated September 30, 2026. It reviewed a set of 12 primary sources in reaching its determination on the claim.
The facts above are stated as recorded in the primary documentation; the exact nature of the safety threshold not met by GPT-6.1 Astra and the potential consequences of its release remain subject to the details contained in those published materials.
The claim that OpenAI is pushing forward with the launch of GPT-6.1 Astra despite the model failing to pass its safety tests is True.
