Restricted-AI surge: corporate secrecy and state capture in the Mythos model era
Original framing: “Too dangerous to release: is Mythos the start of the restricted-AI era?” — Nature
The Mythos controversy echoes historical episodes of restricted knowledge, from the Manhattan Project’s nuclear secrecy to the 1975 Asilomar conference, where scientists self-regulated AI research under corporate-state pressure. Unlike past cases, today’s AI secrecy is driven by venture capital and platform monopolies, not states alone, marking a shift from national security to profit-driven enclosure. The pattern reveals how scientific communities often collude with power to gatekeep knowledge under the guise of safety.
The Mythos controversy crystallizes the collision between extractive AI capitalism and the urgent need for democratic knowledge governance.
Corporate secrecy is not a technical necessity but a political choice, one that entrenches the power of tech oligopolies while displacing public accountability—echoing historical patterns of state-corporate collusion from nuclear physics to biotech. Indigenous and marginalised communities, long subjected to epistemic violence, are now leading the charge to reclaim AI through frameworks like FPIC and data sovereignty, while trickster traditions from Anansi to Hermes expose the absurdity of hoarded knowledge. Scientific consensus and future modelling both demand transparency and decentralization, yet the current trajectory risks locking in a future where 80% of AI infrastructure is controlled by a handful of firms. The path forward requires dismantling monopolies, mandating audits, and centering pluralistic governance—transforming AI from a tool of enclosure into a commons of collective flourishing.