Holy moly: Sam Altman told Axios in a half-hour interview that AI superintelligence is so close, so mind-bending, so disruptive that America needs a new social contract. - It's on the scale of the Progressive Era in the early 1900s, and the New Deal during the Great Depression. - Altman warns: widespread job loss, cyberattacks, social upheaval, machines man can't control - "soon-to-be-released AI models could enable a world-shaking cyberattack this year. "I think that's totally possible," Altman said. "I suspect in the next year, we will see significant threats we have to mitigate from cyber."
One-page infographic (IAPP, Sept 11, 2024) that visualizes major AI risks—including ‘Security and AI safety’, bias, data/privacy issues—and practical mitigations (red‑teaming, safety‑by‑design, impact assessments). It directly supports Altman’s warnings by illustrating the cybersecurity and safety risks of advanced AI and the governance measures experts say are needed to manage workforce, social, and systemic threats.
Source: International Association of Privacy Professionals (IAPP)
Research Brief
What our analysis found
In a half-hour interview published by Axios on April 6, 2026, OpenAI CEO Sam Altman declared that AI superintelligence is imminent and that America needs a new social contract on the scale of the Progressive Era and the New Deal. Altman warned of widespread job loss, social upheaval, and machines that humans may not be able to control. On the same day, OpenAI released a 13-page policy document titled "Industrial Policy for the Intelligence Age," proposing ideas such as robot taxes, a public wealth fund, and automatic safety-net triggers — framing the proposals as "early and exploratory" but underscoring the company's own expectation of massive societal disruption.
Altman's cyber warning was particularly stark: he said soon-to-be-released AI models could enable a "world-shaking cyberattack this year," adding, "I suspect in the next year, we will see significant threats we have to mitigate from cyber." This claim finds partial backing from institutional research. The UK's National Cyber Security Centre published findings on March 30, 2026, showing that frontier AI models could complete over 50% of steps in simulated multi-stage enterprise attacks at a cost of roughly £65 per attempt. NIST released a draft Cybersecurity Framework Profile for AI in December 2025 that explicitly lists "thwarting AI-enabled cyber attacks" as a focus area, while RAND's Forecasting Initiative opened a public question on whether an AI-enabled cyberattack would disrupt critical infrastructure in a G20 country before July 1, 2026.
On the economic front, McKinsey research from November 2025 estimates that technologies demonstrated by 2025 could in principle automate roughly 57% of U.S. work hours, though actual displacement depends heavily on adoption speed, timing, and policy responses. OECD and other major institutions show wide variation in short-term versus long-term job displacement projections, suggesting that while Altman's warning of widespread job loss is directionally supported, the timeline and severity remain deeply uncertain.
Fact Check
Evidence from both sides
Supporting Evidence
Direct Axios reporting confirms the claims
Axios published the half-hour interview on April 6, 2026, with direct quotes from Altman calling for a New Deal-scale social contract and warning that "soon-to-be-released AI models" could enable a world-shaking cyberattack this year. The tweet accurately reflects the reported interview.
OpenAI's own policy blueprint validates the urgency
On the same day, OpenAI released a 13-page document proposing robot taxes, a public wealth fund, and containment playbooks for rogue AI — an explicit institutional admission that the company expects large-scale societal disruption requiring major policy intervention.
UK NCSC empirical testing supports the cyber threat warning
NCSC experiments published March 30, 2026 showed frontier AI models could perform many steps of simulated enterprise attacks at a cost of approximately £65 per full attempt, with the best models completing over 50% of attack steps in some scenarios. The agency warned capabilities are improving rapidly, lending concrete evidence to Altman's cyberattack concerns.
U.S. government frameworks treat AI cyber risk as a near-term priority
NIST's December 2025 draft Cyber AI Profile and broader CISA guidance explicitly identify AI-enabled cyberthreats as an urgent focus area, indicating institutional consensus that the risk Altman describes is real and requires active mitigation.
Industry threat intelligence documents increasing AI use by attackers
Reports from IBM X-Force, CrowdStrike, and Ankura throughout late 2025 and early 2026 document growing use of AI tooling by threat actors for phishing, exploit generation, and automation, consistent with Altman's warning about escalating cyber risk.
McKinsey data supports the scale of potential economic disruption
November 2025 McKinsey research found technologies already demonstrated could automate roughly 57% of U.S. work hours, providing an empirical basis for Altman's warnings about widespread job loss even if the timing remains uncertain.
Contradicting Evidence
No AI model has completed a full realistic end-to-end cyberattack
The same NCSC testing that showed AI models completing many attack steps also found that no public model had successfully executed a complete realistic industrial control system attack as of March 2026, suggesting Altman's implication of imminent "world-shaking" AI cyberattacks may overstate current capabilities.
Job displacement timelines are far more uncertain than Altman implies
While McKinsey's 57% automation-potential figure is striking, the OECD and other major institutions show wide variation in projections, with actual displacement depending heavily on adoption speed, sector differences, and policy responses — making "widespread job loss" a plausible concern but not a near-term certainty.
Altman has a strategic interest in framing AI as transformative
As CEO of OpenAI, Altman benefits commercially and politically from positioning AI as an epoch-defining force, which could incentivize overstating both the pace and scale of disruption. His simultaneous release of a policy blueprint suggests a coordinated messaging strategy rather than a disinterested warning.
"Superintelligence" remains loosely defined and contested
AI researchers and institutions disagree significantly on what superintelligence means and when or whether it will arrive. Altman's claim that it is "so close" lacks a specific technical benchmark, and many experts consider such predictions premature given current model limitations.
Historical analogies to the Progressive Era and New Deal may be misleading
Those transformations unfolded over decades in response to specific economic crises and social conditions. Comparing them to a technology still in rapid development risks conflating speculative projections with historically grounded policy needs, potentially distorting the urgency and nature of the response required.
Report an Issue
Found something wrong with this article? Let us know and we'll look into it.