ALL INPredictions
Last updated Nov 29, 2025
Prediction
Chamath
← Back to hostEpisode
Chamath @ 00:20:16Inconclusive
aitech
Within 5–10 years from May 2024, OpenAI’s internal security and operational protocols will resemble those of a highly classified U.S. "three-letter" agency (e.g., strict access controls, compartmentalization, heavy security procedures) more than the open, campus-like culture historically associated with Google’s Googleplex.
in order to be this incredible bastion of like AGI and innovation, I suspect that it's going to look more like a three letter agency in terms of security and protocols in the next 5 or 10 years, than it is going to look like the Googleplex.
Explanation

The prediction’s time window is 5–10 years from May 2024, i.e., roughly May 2029–May 2034. As of the current date (November 30, 2025), only about 1.5 years have passed, so it’s too early to know whether OpenAI will ultimately resemble a U.S. “three‑letter” agency more than a Google‑style open campus.

There is evidence that OpenAI has recently moved toward a more classified‑style security posture:

  • Multiple reports describe OpenAI overhauling security after concerns about foreign espionage and alleged model copying by the Chinese startup DeepSeek, including stricter controls on sensitive information, enhanced staff vetting, and isolated computing environments. (ft.com)
  • OpenAI has implemented “information tenting”, where access to projects is compartmentalized and employees must verify whether colleagues are “in the tent” before discussing work, which is reminiscent of need‑to‑know compartmentalization used in intelligence agencies. (ft.com)
  • The company reportedly now uses biometric access controls (e.g., fingerprint scans) for certain areas, keeps key systems offline/air‑gapped, and applies a deny‑by‑default policy for internet connections—again, measures commonly associated with high‑security government environments. (allaboutai.com)
  • OpenAI has also hired a CISO from Palantir and added retired U.S. Army General (and former NSA director) Paul Nakasone to its board to advise on cybersecurity and defenses, further signaling a shift toward national‑security‑style governance. (ft.com)

These developments are directionally consistent with Chamath’s claim that OpenAI would evolve to look more like a three‑letter agency in terms of security and protocols. However, the prediction is explicitly about OpenAI’s state 5–10 years out, not about short‑term movement in that direction. Culture, governance, and security posture could still change significantly before 2029–2034.

Because the deadline has not yet arrived and there is no definitive evidence about OpenAI’s eventual resemblance to an intelligence agency versus a Googleplex‑style campus, the correct verdict as of now is “inconclusive (too early)”, though early trends mildly support the prediction.