Interviews with current and former employees reveal how OpenAI has become more secretive as it tries to juggle its ideals with the need to finance its research
The AI moonshot was founded in the spirit of transparency. This is the inside story of how competitive pressure eroded that idealism.
MIT Technology ReviewKaren Hao
Context & Ripple Effects
This 2020 MIT Technology Review investigation is the early marker in an arc the later coverage keeps confirming: OpenAI was founded to publish openly, and employee interviews showed competitive pressure already bending that ideal toward secrecy. What followed validated the diagnosis — by late 2023 the tension between mission and money went public in the boardroom crisis over Altman's firing, with investors complaining the company could not explain its own decisions.
Employees absorb the change directly: a culture built on publication and internal candor gives way to information control, which is exactly what the interviewed staff describe as eroding their ability to trust leadership's stated ideals.
Second-order effects
Rivals gain a legitimacy opening — any competitor can differentiate on openness precisely because OpenAI's transparency record is now documented as compromised, turning 'who shares research' into a recruiting and positioning weapon.
Third-order effects
If the pattern holds, frontier labs converge on a closed, capital-intensive structure where safety review and disclosure are governed by competitive timing rather than founding charters — raising the governance question of who audits a lab whose original accountability mechanism was its own openness.
The trend: Frontier AI labs are institutionalizing around closed, profit-driven operations, trading founding-era transparency for the capital and speed competitive research demands.
I spent half a year digging into @OpenAI, the SF-based AI research lab, originally founded by @elonmusk. I started with a few simple questions: Who are they? What are their goals? How do they work? After nearly three dozen interviews, I found so much more. https://www.technologyr…
A critical review of OpenAI ... parts of it sobering, parts sort of positive. Definitely worth reading ... The messy, secretive reality behind OpenAI's bid to save the world - via @techreview https://www.technologyreview.com/ ...
Above anything else, the directors of this company should give themselves an “F” for failing to live by their own charter, with an undue concentration of power in white male leadership. A(G)I requires diversity and inclusion at the highest levels. https://twitter.com/...
The headline is a bit sensationalist, but the content of this article—balancing transparency and risk, gathering investment and racing to be the 1st while keeping to their altruistic mission—illustrates the difficulties of truly “open AI.” https://www.technologyreview.com/ ...
“They are using sophisticated technical practices to try to answer social problems with AI.” -D&S Affiliate @drbrittparis, quoted by @_KarenHao in @techreview https://www.technologyreview.com/ ...
This is interesting, but what struck me was “In fairness, this lack of diversity is typical in AI. ...women accounted for 20% of AI professorships...” In my *psychology* department, women account for 31% of our professors. Absolutely no where near our PhD grad numbers. https://tw…
Here's an in-depth profile by @_KarenHao of one of the darlings of the AI world, @OpenAI—an instructive tale of what happens when idealism collides with reality in Silicon Valley https://www.technologyreview.com/ ...
“There is a misalignment between what the company publicly espouses and how it operates behind closed doors... OpenAI, for all its noble aspirations, is obsessed with maintaining secrecy, protecting its image, and retaining the loyalty of its employees.” https://twitter.com/...
Here's a profile of an AI startup that is also a profile of Silicon Valley itself, and of what can really be going on behind idealistic narratives of “tech for good.” https://www.technologyreview.com/ ...
Interesting read! History of AGI discussion feels like: - 2017: it's 1-2 years away - 2018: it's 5-10 years away - 2019: >10 years away - 2020: even humans are not AGI... let's imitate them (but that's hard too!) Still lots of progress, as Bengio says: “Science is progressive”! h…
What a fascinating microcosm of the challenges of AI: the drive toward AGI, even though we don't realllly know what it means; the earnest desire to be responsible (same), the lure of hype & the techno-solutionism that can't get out of its own way. Great reporting. https://twitter…