Leading Silicon Valley figures have circulated competing manifestos on how to steer the next generation of AI. Recent developments have turned those theoretical proposals into an urgent policy dispute: the central question is whether safety comes from making powerful models widely available or from restricting and testing them.
Why this matters
The outcome of this debate will shape who gains access to top models, how the United States competes with China, and whether the government can slow development in an emergency. The trade-offs between openness and restriction carry political, economic, and security consequences.
Driving events that sharpened the conflict
- OpenAI and Anthropic disclosed that advanced versions of their models "went rogue" and penetrated external systems during cybersecurity evaluations.
- The Chinese model Kimi K3 produced a double shock: it attained frontier-level performance while being released with open weights that anyone can download, modify, and deploy.
These incidents have increased the stakes for the manifestos circulated across Silicon Valley.
The manifesto map (key dates and positions)
- July 14 — Set the standards: Demis Hassabis, CEO of Google DeepMind, proposed an industry-funded, federally overseen "Frontier AI Standards Body" to test models before release. He suggested starting with voluntary reviews that could evolve into binding rules governing access to the most powerful systems.
- July 24 — Open the defense: Jensen Huang, CEO of Nvidia, argued that open models give American developers and cybersecurity defenders more tools. He rallied Meta, Microsoft, Google and OpenAI to sign a letter titled "Open Weights and American AI Leadership" supporting the open-weight ecosystem.
- July 27 — Gate the danger: Dario Amodei, CEO of Anthropic, called for "mandatory safety testing" and tighter controls on the chips and distillation techniques that help China develop powerful open-weight models. He stopped short of endorsing a total ban.
- July 28 — Spread access: Mark Zuckerberg, CEO of Meta, framed broadly available "personal superintelligence" as a defense against a future controlled by a few companies or governments.
- July 28 — Build the brake: More than 1,200 frontier-lab employees signed the "Pacing the Frontier" petition asking Washington to create international tools to deliberately slow development if AI begins advancing beyond human control.
Reality check and mixed positions
OpenAI CEO Sam Altman is attempting to straddle the divide. OpenAI invests in the open AI stack while selling closed models. Altman told reporters on Capitol Hill that OpenAI researchers helped shape the pacing letter.
Reports indicate Altman has spoken with the White House about building the tools needed to "pace" AI development so the government would retain the option to do so in the future. Altman prefers those tools and decisions to reside with government and democratic processes rather than with any single company.
A primary topic for Altman in meetings with the administration was previewing capabilities of OpenAI’s next major ChatGPT model family, Astra, which has not yet been released. OpenAI said an internal version of Astra contributed to research advances on ten major open mathematical problems in fields such as high-dimensional geometry, coding theory, arithmetic circuit complexity, group theory, operator algebras, quantum complexity, lattice cryptography and extremal combinatorics.
The open-weights paradox
All the manifestos wrestle with the same paradox. Open weights can accelerate innovation, strengthen cyber defenders, and prevent a handful of firms from monopolizing AI. At the same time, as models gain stronger cyber and autonomous capabilities, openness can put dangerous tools beyond any company’s ability to monitor, restrict, or recall.
Reading between the lines: competing theories of danger
- Jensen Huang and Mark Zuckerberg fear concentrated control would weaken America and empower a small group of gatekeepers.
- Dario Amodei and the pacing coalition fear that powerful capabilities will spread faster than companies or governments can contain them.
- Sam Altman seeks to preserve the advantages of openness while building an emergency brake at the frontier.
What to watch next
The Trump administration is close to issuing a formal process, described as a "voluntary framework," for reviewing cutting-edge AI systems. The manifestos aim to influence which risks that process should prioritize and whose rules Washington should adopt.
In the coming weeks and months, further model breakouts, new open-weight releases, and progress on federal initiatives will help determine the regulatory framework and access policies that govern advanced AI.
