The two largest frontier AI labs in the world have spent the past two months issuing AI slowdown coordination calls while simultaneously shipping new models at pace, a tension that sits at the heart of both companies’ public positioning ahead of their anticipated stock market listings.
In a paper released last week, Anthropic called for a coordinated ‘slowdown or pause’ in frontier model development across countries, arguing that policy frameworks need time to catch up. ‘Without a global coordination mechanism, companies and governments will have to make difficult decisions about safety while under competitive and geopolitical pressures,’ the paper stated. Anthropic’s CEO Dario Amodei followed that up in a blog post on Wednesday, describing AI as moving at a ‘lightning pace’ while policy is ‘moving very slowly.’
OpenAI’s CEO Sam Altman and chief scientist Jakub Pachocki made a parallel argument in a Monday blog post, calling for the formation of an ‘international organization that helps coordinate leading AI efforts to reduce catastrophic risk.’ They wrote that such a body should slow frontier AI development so that ‘societal resilience, safety, and alignment can keep pace.’
AI Slowdown Coordination Calls Met With a Sprint of New Releases
Both labs have published these warnings while releasing new models. On Tuesday, Anthropic released Claude Fable 5, described as its latest ‘Mythos-class’ model. OpenAI released GPT-5.5 in late April, calling it its ‘smartest and most intuitive’ model to date and saying it can understand what work a user is doing and carry more of that work itself.
The Fable 5 release carries some context that sharpens the governance question considerably. CyberScoop reported that earlier this year, Anthropic executives said their new AI model, Claude Mythos, had such powerful capabilities for harm that they would not release it publicly. Fable 5 is, in effect, a version of Mythos that has been wrapped in guardrails judged sufficient for public access, a distinction that matters when evaluating the gap between what labs build and what they choose to deploy.
Those guardrails are substantive. According to TechCrunch, in high-risk areas including cybersecurity, biology, chemistry, and distillation attacks, Fable 5 blocks responses and falls back to Claude Opus 4.8 rather than attempting to answer. The report notes that if Fable 5 detects requests related to cybersecurity or distillation attacks, it will either route them to Opus 4.8 or steer users toward a different prompt. The routing mechanism is an architectural safety choice, not merely a content filter applied after generation.
On raw capability, the gap between Fable 5 and the model it routes to is measurable. Lushbinary reports that Anthropic says Fable 5 shows exceptional performance across software engineering and knowledge work, scoring more than 10% higher than Opus 4.8 on some benchmarks. That margin is part of what makes the fallback routing credible as a safety mechanism: Opus 4.8 is meaningfully less capable in the domains where risk is highest, so redirecting sensitive queries there reduces exposure without simply blocking users entirely.
IPO Filings Add a Commercial Layer to the Debate
The timing of the AI slowdown coordination calls has not gone unnoticed. Both companies have filed for public listings: OpenAI said on Monday that it had confidentially filed its S-1, about a week after Anthropic did the same. Public market investors will be reading both the cautionary policy papers and the capability benchmark numbers as part of the same prospectus picture.
Both companies have also been encouraging rapid adoption through free usage perks for their respective developer tools, Codex and Claude Code, through which companies can build AI agents to automate workflows. That push for adoption sits alongside, rather than against, the calls for international oversight, a combination that reflects how competitive pressure and safety advocacy are currently running in parallel rather than in sequence.
Representatives for Anthropic and OpenAI did not respond to requests for comment from Business Insider. OpenAI’s confidential S-1 filing and Anthropic’s own filing mean both companies will soon face the formal disclosure requirements that come with public listings, adding a new layer of accountability to claims made in policy papers and blog posts alike.