Peter Diamandis, founder and executive chairman of the XPRIZE Foundation, has publicly argued that humans behave better when they know they are being observed — a position that places him alongside Oracle co-founder and CEO Larry Ellison in what is becoming a recognizable strand of AI-era techno-optimism about surveillance.

Diamandis's comments arrive roughly two years after Ellison made his own case for AI-powered monitoring at Oracle's 2024 Financial Analyst Meeting. There, Ellison predicted a future in which citizens and police alike would be under continuous, mutual surveillance, characterizing the resulting dynamic as a modern surveillance state. His precise framing, as reported by Fortune, was that such a system would ensure citizens "will be on their best behavior" — a phrase now finding an echo in Diamandis's own articulation of the idea.

The behavioral premise both men are working from is not novel. It has roots in Jeremy Bentham's panopticon concept and, more recently, in empirical criminology literature on deterrence through perceived observation. What is new is the technical substrate: large-scale AI inference pipelines capable of processing video feeds, biometric signals, and behavioral patterns in near real time, at a cost point that makes city-scale deployment plausible. The argument being made by Ellison and now Diamandis is essentially that this substrate finally makes the old behavioral theory actionable.

That framing deserves scrutiny. The claim that surveillance improves behavior rests on a model of human conduct — rational, anticipatory, responsive to monitoring — that holds in some contexts and fails in others. It also sidesteps the question of whose definition of "best behavior" gets encoded into the system and who controls the output. Ellison's 2024 remarks did not engage with those questions in any detail, and if Diamandis's comments follow the same pattern, the gap between the behavioral thesis and the governance architecture required to support it safely remains wide.

Worth flagging: neither Ellison nor Diamandis occupies a neutral position here. Oracle sells cloud infrastructure and database technology that underpins many enterprise and government data systems; Ellison has a direct commercial interest in the normalization of large-scale, AI-integrated data collection. Diamandis, through XPRIZE and his affiliated ventures, has consistently promoted an abundance-and-technology framing of civilizational challenges. Both men are genuine believers in the constructive power of technology, and that belief is on record across decades. But it shapes which costs they foreground and which they elide.

The civil liberties dimensions are well-mapped by this point: function creep in surveillance infrastructure, the chilling effect on lawful dissent, differential error rates in AI-based identification across demographic groups, and the question of legal due process when algorithmic flags precede human review. These are not speculative concerns. They are documented in deployments from London's CCTV network to municipal facial recognition pilots across the United States and China.

What the Diamandis–Ellison convergence does clarify is that a specific techno-philanthropic worldview — one that treats the friction of unwatched human behavior as a problem to be engineered away — is gaining coherent, well-funded advocacy. That is worth tracking independently of whether the underlying behavioral claim is empirically sound.

The harder question, and the one that will determine whether AI-powered surveillance becomes infrastructure or remains contested technology, is regulatory. The EU AI Act classifies real-time remote biometric identification in public spaces as high-risk and bans most such applications outright; the United States has no equivalent federal framework. That asymmetry means the debate Ellison opened in 2024 and Diamandis is now extending will play out differently across jurisdictions — and that the behavioral optimism of both men may run directly into legal architecture designed with different behavioral assumptions about state power.

Technology has repeatedly compressed the cost of observation to the point where previously theoretical capabilities become operational realities. The question now is not whether pervasive AI surveillance is technically achievable. It clearly is. The question is what governance structures get built before it normalizes — and whether the people advocating most loudly for its benefits have thought carefully enough about who sets the parameters.