Sources: Elon Musk shut off OpenAI's access to Twitter's data in December 2022, believing that the $2M/year licensing fee that OpenAI was paying wasn't enough
Context & Ripple Effects
The December 2022 cutoff is one more turn in a long estrangement. Sources reported Musk cut ties with OpenAI back in 2018 after his offer to run it was rejected, and by 2020 he was publicly complaining that exclusive GPT-3 licensing to Microsoft left OpenAI "essentially captured" — the captured-by-Microsoft complaint framing the lab as no longer the open project he backed. The $2M/year Twitter data fee was, in his view, priced for a nonprofit-era relationship that no longer existed.
The move also fits a pattern at Twitter itself: music-licensing talks with the major labels stalled and died after his takeover over the same concern — paying market rate for content the company already held. Musk revived the grievance formally in 2024, renewing his California lawsuit alleging OpenAI breached its founding contract.
First-order effects
- OpenAI loses a real-time social data feed mid-ChatGPT's rise, and its remaining access to Twitter-scale text now runs through Musk's goodwill rather than a $2M contract.
- Twitter's firehose is repriced from a rounding error to a strategic asset under Musk's personal control, with the $2M figure now the public floor he rejected.
Second-order effects
- Other platform owners watching the cutoff have a fresh data point for renegotiating AI-lab licensing deals upward or shutting them down, since Twitter demonstrated access can be revoked unilaterally at will.
- OpenAI's deepening Microsoft relationship — the very one Musk denounced in 2020 — becomes the compensating channel for data and distribution, hardening the capture he criticized.
Third-order effects
- Social platforms' archives shift from sellable inventory to leverage in the AI arms race, with owners like Musk treating data access as a control point over who can build competitive models.
- If access revocation becomes standard practice, model developers are pushed toward owned or licensed corpora and synthetic data, restructuring the data-supply side of the industry around exclusives rather than cheap API deals.
The trend: Platform data access is shifting from cheap bulk licensing to a lever of control in AI competition, with owners repricing or revoking access as models become strategically valuable.