Teahose.
SIGN IN
NEW HERE — WHAT TEAHOSE DOES
We read the entire AI & tech firehose — so you don't have to.
PODPodcastsAll-In, No Priors, Acquired…
NEWNewslettersStratechery, Newcomer…
PAPPapersPhysical AI research
PHProduct Huntdaily launches
VCInvestor ScoutSequoia, a16z, Benchmark…
CLAUDE DISTILLS →
7 reads, 30 sec each — free, 6 AM ET.
+ a live graph of the companies, people & themes underneath.
HOME/AXIOS AI+/📖 Trump's new playbook
NEWS
// NEWSLETTER ISSUE
AXIOS AI+

📖 Trump's new playbook

DATE October 6, 2026SOURCE AXIOS AI+PARTICIPANTS AXIOS AI+
AI SafetyRegulation & PolicyGeopoliticsCompetitive StrategyDefense TechLeadership & Management
In this episode
AnthropicOpenAIGoogleMetaMicrosoftAppleMistral AIMuseReflection AIInstinctCommerce DepartmentU.S. Office of Personnel ManagementGods Don't Give GiftsDavid SacksScott BessentEmil MichaelMisha LaskinJay ClaytonBryan JohnsonScott KuporAndrew N. FergusonSusie WilesPierre Stock
// SUMMARY

1. Key Themes

Theme 1: The Trump administration is betting on flexibility and corporate self-policing over new AI legislation

Static rules risk becoming obsolete

Task force co-chair Scott Kupor argues that locking in a rulebook now could backfire as the technology evolves.

  • "The worst thing for us to do would be to assume a static set of knowledge that allows us to put a regulatory schema in place that then has unintended consequences," said Kupor.
  • On the timeline: "the last thing we want to do is go into a hole for two years and come up with some big policy which will be outdated."

Securities law as a substitute for government oversight

Officials believe market and legal mechanisms can do the work of traditional regulation.

  • "The administration is leaning hard into corporate self-policing to manage the risks. Kupor and other officials believe that third-party audits and direct reporting to corporate boards carry severe securities law implications that rival traditional government oversight."

Regulation may be split across agencies

The likely structural outcome is distributed oversight rather than a single new AI law or regulator.

  • "Kupor suggested one outcome may be that AI regulation is divided among different federal agencies, although it is too early to know what the group — which will do its work over 120 days — will ultimately recommend to the president."
  • "His primary area of focus initially will be to seek input from the energy, manufacturing, security and financial services industries."

Theme 2: AI policy is being forged in the crossfire of competing internal camps

The pro-oversight vs. pro-speed split

The task force pulls together officials with opposing instincts about whether oversight helps or hurts U.S. competitiveness.

  • "Treasury Secretary Scott Bessent and Trump Chief of Staff Susie Wiles have worked behind the scenes to encourage some oversight, including through the voluntary model-review process that already exists."
  • "On the other end of the spectrum, figures like David Sacks and the Pentagon's Emil Michael represent an approach that is wary of the potential for a slowdown in model development or new regulation to reduce competition for frontier companies."

A de facto review regime already exists

Frontier labs are already submitting models for government review, with no new law required.

Cohesion through a central coordinator

Jay Clayton is positioned as the unifier of the competing camps.

  • "Jay Clayton, the director of national intelligence and the new AI czar, is expected to bring cohesion to the group's collective work, much as he does with the myriad intelligence agencies."

Theme 3: Western open-weight models are mounting a comeback against China, with "own vs. rent" as the pitch

New Western flagship open-weight models narrow China's lead

Reflection and Mistral are releasing large mixture-of-experts models aimed at competing with Chinese open models.

  • "A new crop of powerful, Western open-weight AI systems are mounting a comeback, with fresh models from Reflection and Mistral narrowing China's recent lead."
  • "Reflection yesterday introduced Beam, a 501-billion-parameter mixture-of-experts model, with 23 billion parameters active at a time."
  • "Mistral said today it is finishing work on a new flagship model, Mistral Large 4, aka 'Le Chonk.' It's a 1-trillion-parameter multimodal model with 49 billion active parameters."

Ownership and control as the selling point

Both companies position user control, not just raw capability, as the differentiator.

  • "Both companies argue that raw intelligence isn't the only way to measure the AI race, and they present the notion of user control as a core opportunity and selling point."
  • "Once you're spending that amount on intelligence, you want to move from renting it to owning it yourself," said Reflection CEO Misha Laskin.

A third option for enterprises and governments

Open-weight Western models give buyers an alternative to both closed U.S. labs and Chinese options.

  • "These models and others will give businesses and governments alternatives to the closed ecosystems of OpenAI and Anthropic, as well as to Chinese options they may have eschewed for security reasons."

2. Contrarian Perspectives

1. "No new legislation" is a deliberate strategy, not negligence The prevailing criticism is that the administration is doing nothing on AI. Kupor flips it: the absence of a big rulebook is the point.

  • "I reject that concept. We have not come in and said: 'There shall be a new piece of legislation.' And so people think that's asleep at the switch."
  • The supporting logic is that rigid rules get outdated fast, and that existing mechanisms (securities law, board reporting, third-party audits) already carry teeth. The article itself flags the open question: "it remains to be seen whether streamlining existing efforts without new legislation will satisfy Americans who are increasingly anxious about AI safety."

2. Securities law as a regulatory substitute The non-obvious claim is that corporate disclosure and audit obligations can match government oversight in force.

  • "Kupor and other officials believe that third-party audits and direct reporting to corporate boards carry severe securities law implications that rival traditional government oversight."
  • The implication for investors: AI risk governance may increasingly show up as board-level and disclosure obligations rather than agency rules.

3. Raw intelligence isn't the only scoreboard The open-model labs reject the frontier-capability race as the sole metric, arguing that efficiency and control matter.

  • "Both companies argue that raw intelligence isn't the only way to measure the AI race, and they present the notion of user control as a core opportunity and selling point."
  • Evidence: "Reflection says Beam is competitive with China's GLM-5.2 and approaches Alibaba's Qwen 3.8-Max on some coding and agentic tasks, but it uses far less computing firepower."
  • A counterweight risk the article flags: "putting powerful model weights into the world also makes their safeguards easier to remove, a concern that's growing as models become more capable in areas like cybersecurity."

3. Companies Identified

Reflection

  • Description: Western AI lab building open-weight models.
  • Why mentioned: Launched Beam, a 501B-parameter MoE model, as a competitor to Chinese open models.
  • Quotes: "Reflection says Beam is competitive with China's GLM-5.2 and approaches Alibaba's Qwen 3.8-Max on some coding and agentic tasks, but it uses far less computing firepower." / "Reflection plans to release Beam's weights later this month, along with tools for running, evaluating and fine-tuning it."

Mistral

  • Description: European AI lab building open-weight models.
  • Why mentioned: Finishing Mistral Large 4 ("Le Chonk"), a 1T-parameter multimodal model it claims is the best open-weight model in the world.
  • Quotes: "Stock told Axios the company believes Le Chonk is the world's best open-weight model." / "It plans to release the weights Oct. 27 after further reinforcement learning and safety testing."

Anthropic

  • Description: Frontier AI lab behind Claude.
  • Why mentioned: Participates in the voluntary government model-review process; also cited as a closed ecosystem that open models compete with; and Meta and Microsoft are trying to reduce employee use of its Claude.
  • Quotes: "the reviews have been used for top models at Anthropic, OpenAI and now Google" / "Meta and Microsoft are working to cut their employees' use of Anthropic's Claude."

OpenAI

  • Description: Frontier AI lab.
  • Why mentioned: Participates in the Commerce-managed model reviews; cited as a closed ecosystem.
  • Quotes: "the reviews have been used for top models at Anthropic, OpenAI and now Google" / "alternatives to the closed ecosystems of OpenAI and Anthropic"

Google

  • Description: Big tech and frontier model developer.
  • Why mentioned: Newest addition to the voluntary model-review process.
  • Quotes: "the reviews have been used for top models at Anthropic, OpenAI and now Google, sources tell Axios."

Meta and Microsoft

  • Description: Big tech companies and heavy enterprise AI users.
  • Why mentioned: Reportedly working to cut employees' use of Anthropic's Claude.
  • Quotes: "Meta and Microsoft are working to cut their employees' use of Anthropic's Claude. (The Information)"

Apple

  • Description: Consumer tech giant operating the App Store.
  • Why mentioned: Its highly profitable App Store revenue is threatened by personal AI assistants.
  • Quotes: "The rapid growth of personal AI assistants, like Instinct and Muse, is a threat to Apple's highly profitable App Store revenue."

Instinct and Muse

  • Description: Personal AI assistant products.
  • Why mentioned: Cited as examples of fast-growing assistants that could disintermediate the App Store.
  • Quotes: "The rapid growth of personal AI assistants, like Instinct and Muse, is a threat to Apple's highly profitable App Store revenue."

Alibaba (Qwen) and GLM (China's open models)

  • Description: Chinese open-weight model developers.
  • Why mentioned: The benchmark Western open models are measured against.
  • Quotes: "Reflection says Beam is competitive with China's GLM-5.2 and approaches Alibaba's Qwen 3.8-Max on some coding and agentic tasks."

"Gods Don't Give Gifts"

  • Description: A film billed as the first fully AI-made feature.
  • Why mentioned: Being submitted for a best animated feature Oscar.
  • Quotes: "The creators behind 'Gods Don't Give Gifts' are submitting the film for a best animated feature Oscar."

4. People Identified

Scott Kupor

  • Description: Silicon Valley veteran; director of the U.S. Office of Personnel Management; co-chair of the "super intelligence" task force.
  • Why mentioned: Main voice defending the flexible, no-new-legislation approach.
  • Quotes: "Kupor, a Silicon Valley veteran who is currently serving as director of the U.S. Office of Personnel Management, akin to an HR director for the federal government." / "At the end of the day there is no daylight between the president and anyone else ... we need to win."

Jay Clayton

  • Description: Director of national intelligence; the new AI czar.
  • Why mentioned: Expected to unify the competing factions.
  • Quotes: "Jay Clayton ... is expected to bring cohesion to the group's collective work, much as he does with the myriad intelligence agencies."

Emil Michael

  • Description: Pentagon undersecretary; task force co-chair.
  • Why mentioned: Represents the camp wary of regulation slowing frontier development.
  • Quotes: "figures like David Sacks and the Pentagon's Emil Michael represent an approach that is wary of the potential for a slowdown in model development or new regulation to reduce competition for frontier companies."

Andrew Ferguson

  • Description: Federal Trade Commission chair; task force co-chair.
  • Why mentioned: Named co-chair of the task force.
  • Quotes: "Federal Trade Commission Chair Andrew Ferguson and Pentagon undersecretary Emil Michael will also co-chair the task force."

David Sacks

  • Description: Administration figure aligned with the pro-speed, anti-slowdown camp.
  • Why mentioned: Representative of the skeptics of new regulation.
  • Quotes: "figures like David Sacks and the Pentagon's Emil Michael represent an approach that is wary of the potential for a slowdown in model development or new regulation..."

Scott Bessent and Susie Wiles

  • Description: Treasury Secretary and Trump's Chief of Staff.
  • Why mentioned: Behind-the-scenes advocates for some oversight, including the voluntary model-review process.
  • Quotes: "Treasury Secretary Scott Bessent and Trump Chief of Staff Susie Wiles have worked behind the scenes to encourage some oversight."

Misha Laskin

  • Description: CEO of Reflection.
  • Why mentioned: Articulated the "rent vs. own" thesis for open-weight models.
  • Quotes: "Once you're spending that amount on intelligence, you want to move from renting it to owning it yourself."

Pierre Stock

  • Description: VP of science at Mistral.
  • Why mentioned: Voiced the anti-oligopoly rationale and the claim that Le Chonk is the best open-weight model.
  • Quotes: "I don't want to live in the future in which any oligopoly controls closed access to this type of intelligence."

Bryan Johnson

  • Description: Longevity influencer.
  • Why mentioned: Interviewed by the newsletter author for "The Axios Show."
  • Quotes: "I just interviewed longevity influencer Bryan Johnson for 'The Axios Show.'"

5. Operating Insights

1. Pitch ownership and control, not just benchmark performance. Reflection and Mistral are differentiating on user control and cost rather than leaderboard dominance. For founders selling AI infrastructure or applications, "move from renting to owning" is a framing that resonates once customers' spend scales.

  • "Once you're spending that amount on intelligence, you want to move from renting it to owning it yourself."

2. Expect governance expectations to flow through boards and audits. If the administration's self-policing approach holds, AI companies and their customers should anticipate scrutiny via third-party audits and board-level reporting, with securities-law exposure attached. Building audit-ready processes now is a hedge.

  • "third-party audits and direct reporting to corporate boards carry severe securities law implications that rival traditional government oversight."

3. Plan for staged, safety-tested releases with tiered access. Mistral's rollout model (moderated API first, a less restricted variant for select partners, public weights later after safety testing) shows a way to ship powerful models while managing misuse risk.

  • "Mistral is initially making the model available through a moderated API, with a version with fewer restrictions and broader cybersecurity capabilities being shared with select partners for testing. It plans to release the weights Oct. 27 after further reinforcement learning and safety testing."

6. Overlooked Insights

1. Big tech is quietly de-risking dependence on a single model vendor. A brief mention suggests large enterprises are managing concentration risk in their AI tool stack.

  • "Meta and Microsoft are working to cut their employees' use of Anthropic's Claude."

2. The task force's 120-day clock and industry-first input sequence. The short timeline and stated focus on energy, manufacturing, security and financial services signal where policy attention (and potential rules) may land first.

  • "the group — which will do its work over 120 days"; "His primary area of focus initially will be to seek input from the energy, manufacturing, security and financial services industries."