All change (again) at OpenAI
Non-profit board to lose control, more senior execs quit
The best-known generative AI company is abandoning its former corporate structure, allowing lawyers to see its training data, and shedding more top executives.
OpenAI abandons non-profit status
OpenAI is to become an independent for-profit public benefit corporation, with its non-profit board losing control of the for-profit company, according to a report by Reuters.
The non-profit board's previous structure, with its members keeping a firm hand on the multi-billion-dollar company, to ensure that it stuck to its stated mission of creating "safe AGI that is broadly beneficial".
Last November, the non-profit board attempted to oust CEO Sam Altman, but Microsoft, a major investor, threatened to hire Altman and key staff, removing them from the company, after which he was reinstated.
Investors have chafed at the presence of the non-profit board, and its removal would make OpenAI’s structure more like that of rival firms such as Anthropic. The exact nature of the changes and the timeline are still be worked out, according to Reuters, which cited unnamed sources.
Assuming the measures go through, Altman will receive a large amount of equity, and current employees will be able to cash out their "pseudo-shares" into the ongoing $6.5 billion funding round.
The non-profit board will remain a minority shareholder in the for-profit company, but its role will be limited.
CTO Mira Murati quits
In yet another change at the top at OpenAI, CTO Mira Murati, has quit the company.
Murati was a key figure in the development of the company's ChatGPT and DALL-E models.
Two more executives, chief research officer Bob McGrew and research vice president Barret Zoph, have also announced their plans to leave.
These departures are the latest in a string of high-profile executive exits from OpenAI in recent months, including Chief scientist and superalignment lead Ilya Sutskever, who left in May.
Lawyers to be allowed to see OpenAI's training data - under strict conditions
OpenAI has agreed to reveal the data used to train its generative AI models to US attorneys pursuing copyright claims on behalf of several authors.
The authors sued OpenAI last year, arguing that its AI models have been trained on their books in violation of US copyright law.
The terms of access to the training data are extremely strict. "Training Data shall be made available by OpenAI in a secure room on a secured computer without Internet access or network access to other unauthorized computers or devices," an order by judge Robert Illman states. T
OpenAI has not responded to requests for an explanation of the secrecy, but it is likely due to fear of legal liability and the potential for more lawsuits.
New AI regulations, such as The EU AI Act and California's AI data transparency bill, include measures to increase the transparency around data used to train AI models.
OpenAI's legal team has argued that generative AI is about creating new content and that the processing of copyrighted works during model training does not infringe copyright.