Editorial illustration for Anthropic co-founder says singularity underway amid AI security focus
Anthropic Co-Founder: AI Singularity Already Underway
Jack Clark didn't mince words at the IMF and World Bank's annual meetings this week. The Anthropic co-founder told the room of economists and policymakers that the singularity, the point where AI starts outpacing human ability to track its own progress, has already started. Not coming.
Not close. Underway.
His evidence leans on test scores most people have never heard of. Humanity's Last Exam, a set of expert-level questions designed to stump AI models, saw scores jump from 8% to 67.7% since 2025. FrontierMath, a brutal math benchmark, went from 2% to 94% in roughly the same window. Clark used those numbers to argue something structural has shifted in how fast AI systems are improving, not just incrementally but at a pace that outruns normal human oversight.
That framing matters because Clark didn't stop at benchmarks. He tied the acceleration to economic forecasts, labor disruption, and a question Anthropic CEO Dario Amodei has reportedly raised: whether AI's intelligence needs a ceiling. What Clark said next is where it gets specific.
Singularity talk used to live strictly in science fiction, and now one of the people building it is saying it out loud to the folks who run the global economy.
Why this matters
Clark isn't some outside commentator hyping AI, he's co-founding a company that just spent a year telling customers to slow down and build guardrails. That contradiction is worth sitting with. If Anthropic's own leadership believes the singularity has already started, the $44 trillion economy number and the talk of an intelligence cap aren't thought experiments anymore, they're policy questions with a shrinking deadline attached.
Meanwhile NorForge's finding that a Chinese-speaking hacker let AI agents handle nearly all the attack work shows the gap between "AI moving fast" and "AI security keeping up" is already a live problem, not a future one. For developers and founders, the practical read is this: the tooling for managing AI output and building an AI-native SDLC matters more than ever, especially with only 23% of leaders reporting an actual AI strategy. Rhetoric about singularities is cheap.
What's not cheap is the fact that offense is outpacing defense right now, and most organizations are still unprepared.
Related Reading
Browse all Anthropic news →
Business & Startups Nordic pilot adds Gemini for Education, NotebookLM to boost AI literacy
Business & Startups Kling launches Video O1, all-in-one model with MVL bridge using transformer
Business & Startups DeepSeek Seeks More Capital Weeks After USD 7B Funding Round
Related Topic Claude gains shared context in Excel, PowerPoint; Microsoft adds Copilot Cowork
AI Tools & Apps
Related Topic Anthropic: Hackers Stole Claude Tokens Via Compromised Session Key
LLMs & Generative AI
From Archives Microsoft CEO Nadella warns AI agents are like insider threats
From Archives Nvidia Launches Open Models for Enterprise AI Agents
From Archives Anthropic, Google block OpenCode API keys after hackers ran up user charges
LLMs & Generative AI
From Archives Anthropic cushions IPO with fund for workers its models might replace
Common Questions Answered
What evidence does Jack Clark cite to support his claim that the singularity is already underway?
Jack Clark points to test scores from Humanity's Last Exam, a set of expert-level questions designed to challenge AI models, which showed dramatic improvement from 8% to 67.7% in performance. This benchmark demonstrates that AI systems are rapidly closing the gap on tasks that were previously considered difficult for artificial intelligence to master.
Why is Jack Clark's singularity claim particularly significant coming from an Anthropic co-founder?
Clark's statement carries weight because Anthropic spent the previous year advising customers to slow down AI development and build safety guardrails, creating an apparent contradiction with his singularity declaration. This suggests that even the leadership of a company focused on AI security believes we have already reached a critical inflection point in AI capability advancement.
What does Jack Clark mean by the singularity being 'underway' rather than 'coming'?
Clark's distinction emphasizes that the singularity—the point where AI begins outpacing human ability to track its own progress—is not a future event but is currently happening. His statement to economists and policymakers at the IMF and World Bank annual meetings suggests that this transition has already commenced, not that it remains a distant possibility.
How do Clark's comments about the singularity affect policy discussions according to the article?
If Anthropic's leadership genuinely believes the singularity has started, discussions about economic impacts and intelligence caps transform from theoretical thought experiments into urgent policy questions with time-sensitive implications. This shift in perspective suggests policymakers need to act quickly on AI governance and regulation rather than treating these concerns as distant future concerns.
Further Reading
- AI rapidly improving, could vastly boost US GDP, Anthropic's Clark tells policymakers - Reuters
- Anthropic co-founder warns AI's window for oversight is narrowing - The Daily Star
- Anthropic co-founder renews call for robust regulation of frontier AI - Malay Mail
- Hope or Threat? Anthropic Cofounder on AI's Future - NHK World
- Import AI 477: In which I tell the IMF the Singularity has begun - Import AI
