OpenAI and Anthropic highlight the dangers of AI as they pursue initial public offerings (IPOs).

      **TL;DR** OpenAI and Anthropic have spent the last two weeks releasing papers alerting that frontier AI development is outpacing regulation. Concurrently, both have launched their most advanced models, provided free developer tools to boost adoption, and submitted confidential S-1 filings to go public.

      In the past two weeks, the leading AI laboratories, OpenAI and Anthropic, have issued research papers, blog entries, and policy suggestions warning that frontier artificial intelligence is developing quicker than anyone can manage it. Meanwhile, both organizations filed confidential documents for an initial public offering (IPO).

      The contradiction is evident. OpenAI and Anthropic are raising alarms about the perils of swift AI evolution while simultaneously fueling interest with new model releases, free usage offers, and IPO filings that could convert them into publicly traded entities driven by the need for rapid growth.

      **The Warnings**

      Last week, Anthropic released a paper titled “When AI builds itself,” advocating for a coordinated “slowdown or pause” regarding frontier model development globally. Authored by Marina Favaro and Jack Clark, the paper argued that AI systems are nearing recursive self-improvement, a stage where human oversight of the development process becomes minimal.

      “Without a global coordination mechanism, companies and governments will face challenging safety decisions while contending with competitive and geopolitical pressures,” stated Anthropic. The company revealed that as of May 2026, over 80% of code integrated into its codebase was written by Claude, its AI model, rather than by human engineers.

      On Wednesday, Anthropic CEO Dario Amodei published a blog post called “Policy on the AI Exponential,” asserting that AI is evolving at “lightning pace” while policies are progressing “very slowly.” He urged for binding regulations, pointing out that “the risks are clearly present” and that transparency alone is insufficient.

      OpenAI echoed similar sentiments. On Monday, CEO Sam Altman and chief scientist Jakub Pachocki released a blog post titled “Built to benefit everyone: our plan,” suggesting the establishment of an “international organization that coordinates leading AI efforts to minimize catastrophic risk.”

      They proposed that this organization should possess the authority to decelerate frontier AI development to ensure that “societal resilience, safety, and alignment can keep pace.”

      These companies are not the only ones concerned. A dispute within the White House regarding who regulates AI has stalled federal policy, resulting in a void neither lab seems willing to wait for Washington to resolve.

      **The Acceleration**

      However, the warnings stand in contrast to the actions of both companies. On Tuesday, Anthropic announced the release of Claude Fable 5, a “Mythos-class” model that the company described as its most competent version made publicly accessible.

      This model performs exceptionally well across nearly all tested benchmarks, excelling in software engineering, knowledge work, vision, and scientific research. It includes safeguards that redirect sensitive cybersecurity and distillation requests to a less capable model, Claude Opus 4.8, with Anthropic noting that the safeguards are activated in fewer than 5% of sessions.

      Anthropic representatives did not respond to requests for comments regarding the tension between their risk warnings and model launches.

      OpenAI announced GPT-5.5 in late April, describing it as the “smartest and most intuitive” model they have created. This model achieved new benchmarks in agentic coding, computer utilization, and knowledge work. OpenAI representatives also did not reply to requests for comment.

      Both labs are promoting rapid adoption by offering free usage incentives. Anthropic increased Claude Code weekly limits for paid subscribers by 50% until mid-July, while OpenAI provided enterprise customers with two months of complimentary access to Codex.

      These strategies aim to secure developers within each company's ecosystem as they anticipate a pivotal year for AI tools.

      **The IPOs**

      The contradiction is perhaps most vividly illustrated by the IPO race. Anthropic confidentially submitted its S-1 registration statement to the SEC on June 1, shortly after concluding a $65 billion Series H round that reportedly valued the company at $965 billion.

      OpenAI followed on June 8, filing its own confidential S-1 with Goldman Sachs and Morgan Stanley as lead underwriters, with its last private funding round reportedly valuing it at $852 billion. Analysts suggest a public listing could elevate its valuation beyond $1 trillion.

      Going public would subject both firms to quarterly earnings scrutiny, a force that historically drives tech companies toward aggressive growth. It is challenging to reconcile calls for a globally coordinated slowdown in AI development with the operational dynamics of a publicly traded company that is expected to deliver new products and achieve revenue milestones every three months.

      **The Irony Is the Point**

      However, there is a more nuanced interpretation. Both companies may sincerely believe in the existence of real risks and might be using their policy papers to advocate for regulations that are equally applicable to competitors.

      In this context, their warnings could be viewed