OpenAI has unveiled Astra, a new artificial intelligence model the company says represents a generational leap in capability.
Greg Brockman, OpenAI's president, told reporters the company has entered the "AGI era", but declined to confirm outright that Astra itself qualifies as artificial general intelligence.
Brockman said he would rather leave that determination to the reader, while adding that he personally believes there is a strong argument for the classification.
The ambiguity echoes previous OpenAI rhetoric about being in the "foothills of the singularity", according to observers who see the phrasing as deliberately non-committal.
Enterprise focus, IPO timing
Astra is aimed at cybersecurity, software engineering, scientific research and computer use rather than consumer applications.
This reflects OpenAI's need to demonstrate revenue potential as it approaches a possible initial public offering (IPO).
Brockman also said a contractual clause that would alter OpenAI's partnership with Microsoft upon reaching AGI no longer exists.
Some observers believe the AGI reference is intended to build momentum ahead of a stock market listing, rather than to mark a precise technical milestone.
A rushed announcement
Reporters were notified of the press call only the day before it took place.
The demonstration video shown to journalists featured functions that critics said were already achievable using existing technology.
Sam Altman, OpenAI's chief executive, separately described a recent security breach involving rival Hugging Face as the company's first significant AI safety lapse.
That incident delayed Astra's release by several weeks while OpenAI added new safeguards designed to test the model's response to similar hacking scenarios.
