OpenAI reframes access and safety
- OpenAI on April 26 published a new “Our Principles” statement, with Sam Altman shifting the company’s public framing toward wider AI access and democratic governance. - The document says OpenAI will “resist” power concentrating among a few companies, while still requiring safety problems be solved before proceeding further. - The rewrite updates OpenAI’s 2018 AGI-first language as GPT-5.5 rollouts and public release notes make staged deployment more visible. (openai.com)
OpenAI published a new “Our Principles” statement on April 26, recasting its public mission around broad access, democratic decision-making, and safety checks. (openai.com) In the new document, Sam Altman says future AI power should not sit with “a small handful of companies” and argues for putting “truly general AI” in as many hands as possible. (openai.com) The first principle is “Democratization,” followed by “Empowerment” and “Universal Prosperity.” The text says OpenAI wants users to have broad latitude, while also minimizing catastrophic harm, local harms, and “corrosive societal effects.” (openai.com) A Euronews review published April 27 says the rewrite marks a clear shift from OpenAI’s 2018 principles, which put stronger emphasis on building artificial general intelligence safely and beneficially. (euronews.com) The new version still keeps OpenAI’s core mission statement — ensuring AGI benefits all of humanity — but it places more weight on iterative rollout. The text says society needs to contend with “each successive level of AI capability,” integrate it, and decide the next step together. (openai.com) (euronews.com) OpenAI also says it expects to work with governments, international agencies, and other AGI efforts to solve serious alignment, safety, or societal problems before moving further. Euronews says that language is new in the company’s public principles. (openai.com) (euronews.com) That change lands as OpenAI is still shipping products in public. ChatGPT release notes updated April 23 say GPT-5.5 is rolling out in ChatGPT, with “Fast answers” and a clinician-focused version added on April 22. (help.openai.com) OpenAI’s safety pages now present deployment as a cycle of teaching, testing, and sharing, with red teaming, system cards, preparedness evaluations, safety committees, and feedback listed as part of the process. (openai.com) The company also updated the GPT-5.5 system card on April 24 to add more information about safeguards for GPT-5.5 and GPT-5.5 Pro in the application programming interface, or API. (openai.com) Taken together, the April 26 principles rewrite and the April product notes show OpenAI tying its case for wider access to a more explicit public record of staged deployment and safety controls. (openai.com) (help.openai.com) (openai.com)
