Global AI Pause Urged
Anthropic Urges Global Pause in AI Development Citing Risks of Self-Improvement
Anthropic, the AI company now valued at $965 billion, has called for a worldwide pause in the development of frontier artificial intelligence. The company published a landmark report on June 4 warning that AI is approaching a point where it could improve itself without human input, according to Diario de Ibiza. The report, titled "When AI Builds Itself," was written by Marina Favaro of the Anthropic Institute.
The call for a pause comes just days after Anthropic closed a $65 billion funding round, pushing its total raised to $130 billion since 2021. CEO Dario Amodei warned that "no action is too extreme when the fate of humanity is at stake." Critics, however, say the timing is suspicious — the company is also preparing for a $1 trillion IPO.
AI Writing Its Own Code Triggers the Alarm
The trigger for Anthropic's warning is its own unreleased model, codenamed "Mythos." According to La Opinión de Murcia, 80% of Anthropic's internal code is now written by AI. Engineers are shipping code 8 times faster than they were in 2024. Anthropic warns this creates a feedback loop — AI designing its own successor with less and less human input.
The Mythos model also scored 83.1% on the CyberGym benchmark, a 17-point jump over its predecessor. It identified over 10,000 critical software vulnerabilities through Project Glasswing, a defensive cybersecurity program now expanded to 200 organizations. Anthropic's own data shows the model solves 97% of safety research tasks that take humans a week to reach just 23% completion, according to Levante-EMV.
Anthropic Proposes a Nuclear-Style Global Treaty
Anthropic is not just asking companies to slow down voluntarily. The company wants a treaty similar to the Intermediate-Range Nuclear Forces Treaty, according to El Periódico de Aragón. The plan calls for satellite monitoring of GPU clusters and power grids worldwide. Co-founder Jack Clark put it plainly: "The industry has a gas pedal but no brake."
On June 5, UN Secretary-General António Guterres called for an emergency ministerial session in Geneva. He proposed an international AI regulatory body modeled after the International Atomic Energy Agency. The EU AI Office has also invited Anthropic to brief its Independent Scientific Panel, signaling a possible update to the EU AI Act, according to La Provincia.
Critics Say It Is a Business Move, Not Safety
Not everyone is convinced. Venture capitalist David Sacks called the pause a "performative stunt" designed to hurt open-source competitors before Anthropic's IPO. Analysts at Bloomberg Tech and online communities point out that a pause "locks in" Anthropic's current lead over OpenAI and Google, according to El Periódico de Extremadura.
The Trump administration also pushed back. President Trump signed an executive order on June 2 creating only a voluntary 30-day review period for new AI models. White House officials warned that any hard slowdown would hand a strategic edge to China, according to La Opinión de Málaga. Trump has framed U.S. AI policy as "America First" — fast development, with government access to top models.
A $965 Billion Company Asking the World to Slow Down
The contradiction is hard to miss. Anthropic's annualized revenue crossed $47 billion in June 2026, up from $14 billion in January. The company is on track to hit $50 billion by the end of the month, according to La Opinión de A Coruña. It is racing toward a $1 trillion valuation while asking everyone else to stop racing.
Amodei has also warned that AI could eliminate 50% of white-collar entry-level jobs within a few years. If recursive self-improvement — AI building better versions of itself — arrives within two years as Anthropic predicts, the human role in software engineering could shrink to pure supervision, or disappear entirely, according to Información.
Sentiment
NeutralPublishers
9
Articles
0
Reach
9
