OpenAI ищет тормоза: пост директора по готовности наконец занят

OpenAI appointed a director of readiness, and it sounds like a position from a Hollywood thriller about machine uprising. While Sam Altman sells the world a concept of a future where AI solves all of humanity's problems, inside the company they clearly understand that this future might arrive too quickly and not quite as planned. Appointing someone to a role responsible for safety and "readiness" for risks is not just bureaucratic formality, but an attempt to build a system of checks and balances in an organization often accused of excessive haste.

The context here is more important than the appointment itself. After Ilya Sutskever and his "superalignment" team essentially disappeared from radar following November's coup at OpenAI, a logical question arose: who now watches to ensure AI doesn't teach a schoolkid how to build a bomb in his garage? The answer came in the form of the Preparedness Framework.

This is an internal regulation that describes how the company will evaluate its models across four risk categories: cybersecurity, chemical and biological threats, nuclear safety, and of course, good old "autonomous behavior." The essence of this framework is simple: if a model receives a "High Risk" rating in any of the categories, it cannot be released for public use. The Director of Readiness becomes that very person who holds their hand on the shutoff switch.

They will have to balance between investors' desire to release a product as quickly as possible and the need to guarantee that this product won't crash financial markets or begin manipulating the consciousness of millions of users. Why is this happening right now? Simple: GPT-5 looms on the horizon.

Each new generation of models demonstrates exponential growth in capabilities, and regulators in the US and Europe are no longer willing to take OpenAI at their word. They need specific people responsible for specific processes. It's also an excellent PR move: to show that the company is maturing and ready to play by the rules, even if it writes those rules for itself.

In reality, we see an attempt to legitimize the development of ultra-powerful systems. If you have a "director of readiness," it means you're ready for any scenario, right? At least, that's how it should look in the eyes of the public.

But a question remains open: can one person, even with the most formidable title, stop a multi-billion-dollar machine if it suddenly decides it no longer needs the rules? In an industry where the speed of releasing updates determines survival, safety often becomes an annoying obstacle that one wants to circumvent. The main point: OpenAI is creating a facade of a responsible corporation before launching its next big model.

Will AI become safer because of this, or is it just masterful work by lawyers and the HR department?