David Sacks, the venture capitalist who spent 15 months as the White House’s first AI and crypto czar, used a Saturday post on X to tell Anthropic and OpenAI exactly what he thinks of their joint push to “pace the frontier.” His answer: go ahead, but do it yourselves, and stop asking Washington for anything in exchange. The comments, reported by The Next Web on September 13, 2026, land in the middle of the most heated week of AI-safety politics the industry has seen since the two labs’ chief executives publicly endorsed slowing down model development just days earlier.
Sacks co-chairs the President’s Council of Advisors on Science and Technology – a role TIME reported he stepped into immediately after leaving the AI czar post in March 2026 – and served as the administration’s AI and crypto czar until his 130-day term as a special government employee expired, with Reuters pinning the end of that term to March 2026, even as CBS News later described him stepping down in May 2026 – accounts that still haven’t fully converged on the exact exit date. He remains one of the most influential voices shaping how the White House talks about frontier AI, and his intervention this week reframes a safety debate that Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman had tried to steer toward voluntary restraint.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
What Sacks actually said about Anthropic and OpenAI
According to The Next Web’s report, Sacks told Anthropic and OpenAI to “go ahead and pace the frontier” but to stop seeking an antitrust waiver or an approval regime, arguing that product liability and market pressure already discipline the two companies. On X, he wrote: “Dario has written that we need to ‘pace the frontier,’ and Sam has agreed. People may be surprised by my response: go ahead. You guys are the frontier. By any reasonable metric – market share, revenue growth, model capability – the two of you have a duopoly on frontier,” a post also carried by Benzinga via TradingView.
He went further than a simple shrug. Sacks argued that if the two labs believe their unreleased models are genuinely too dangerous to ship, nothing stops them from delaying release on their own authority. What he rejected was the idea that restraint should come bundled with new rules for everyone else. Companies asking regulators to help them pace the frontier, he wrote, should stop pretending antitrust law must be suspended “so you can form a cartel,” and should stop demanding an approval process that supersedes existing product liability law. The Next Web quoted him summarizing the logic bluntly: “The easiest way not to build superintelligence is for you to agree not to build it.” The framing was consistent with remarks he had made just days earlier: on September 14, 2026, CBS News reported Sacks arguing that responsibility for AI safety belongs with developers rather than regulators, and by September 16, 2026 both Reuters and Politico had quoted him dismissing AI-apocalypse warnings outright, calling them a “fear-mongering” campaign in Reuters’ account and an outright “hoax” in Politico’s.
Sacks also took direct aim at METR, the third-party evaluator Anthropic uses to test frontier models for dangerous capabilities, questioning its independence and suggesting the organization is too intertwined with Anthropic’s own investors and staff to serve as a neutral safety referee. Demanding a regulatory framework as the price of restraint, he added, would look like “blackmail.”
The Amodei-Altman “pace the frontier” pitch that triggered this
Sacks was not responding to nothing. In the days before his post, Anthropic’s Dario Amodei publicly argued that the industry needs to “pace the frontier,” slowing the rate at which model capabilities improve rather than racing at full speed toward more powerful systems. Sam Altman backed him, telling reporters OpenAI would do the same. The Next Web covered that alignment between the two rival CEOs under the headline “Sam Altman backs Dario Amodei’s call to slow down, and says OpenAI will do the same,” framing it as a rare moment of public agreement between companies that otherwise compete fiercely for enterprise customers, developers, and capital.
That is precisely the framing Sacks rejected. His argument treats Amodei and Altman’s companies as a two-player market at the very top of the capability ladder, a “duopoly” in his words, with enough market share, revenue growth, and raw model strength to set the pace for the entire industry without needing anyone’s permission. If they think a model is too risky, he argues, they can simply hold it back. What they cannot do, in his view, is turn that decision into a request for a bespoke regulatory framework that would also bind smaller competitors and open-source developers who never asked for the same protection.
The Anthropic resignation shadowing the whole debate
Sacks’ post did not arrive in a vacuum. It landed squarely inside the fallout from a viral resignation at Anthropic. A pretraining researcher named Jacob Coxon left the company on September 8, 2026, saying publicly that the industry was “gambling with people’s lives,” a departure that kept AI-safety debates in the headlines for days, according to reporting picked up by TradingView and Pakistan’s The News.
Anthropic’s own leadership has not distanced itself from that alarm. Evan Hubinger, who heads Anthropic’s Alignment Science team, has publicly estimated the odds of AI causing human extinction within the next decade at more than 10%, a figure that multiple outlets, including TradingView and The News, cited directly in connection with Coxon’s resignation. That 10% estimate, coming from inside the company rather than from an outside critic, has become one of the most-quoted numbers in this week’s coverage of the standoff.
Sacks did not simply dispute the policy implications of Coxon’s departure. He also questioned the speed and shape of its viral spread, noting that the account behind the resignation thread had apparently posted nothing and had no followers before the thread appeared, and that any earlier activity seemed to have been deleted. Jason Calacanis, Sacks’ co-host on the All-In podcast, added that Coxon had approached the show seeking an interview and then canceled the morning it was scheduled. Sacks suggested the rollout looked coordinated rather than organic, a claim that has itself become a secondary front in the argument over whether Anthropic’s safety warnings are being amplified for strategic reasons.
Why liability, not approval, is the fight underneath the fight
Underneath the personal sparring is a genuine and consequential policy disagreement about how frontier AI should be governed in the United States. Sacks’ preferred model leans on product liability and market discipline rather than a pre-release approval process. His argument: if a model enables a serious cyberattack or otherwise causes harm, the company behind it already faces liability exposure, and markets already punish vendors whose systems behave unpredictably. Layering a new federal approval regime on top of that, in his view, does nothing except hand incumbents a legal shield and a competitive moat.
That is not a new position for Sacks. In June 2026, when Anthropic floated the idea of a mechanism that would let frontier labs pause development if needed, Sacks accused the company on X of running what he called a “regulatory capture scheme,” writing: “Signs you might be trying to get your frontier AI lab nationalized: You compare it to nukes… threaten half of white-collar jobs… warn recursive self-improvement could end humanity… then race ahead anyway.” The following month, he criticized Anthropic’s resistance to releasing open-weight models, writing that “the entire tech industry (save for Anthropic) has come out in favor of open source AI,” and warning that “they won’t stop until they kneecap open source… the rest of the industry needs to watch these guys like a hawk.”
Europe already chose, and chose both
The European Union offers a live test case for the argument Sacks is making, and the comparison cuts in an unexpected direction. The European Commission withdrew its proposed AI Liability Directive in February 2025 after deciding that revised product liability rules already covered AI software adequately. Starting December 9, 2026, those rules treat standalone software, operating systems, and AI systems as products under strict liability. A claimant will not need to prove fault, and “defectiveness” now explicitly covers cybersecurity vulnerabilities and self-learning behavior that turns unsafe after a product has already shipped. Courts will be able to compel a defendant to disclose evidence and can presume a defect exists where a claimant faces excessive difficulty proving one otherwise.
What Europe did not do is treat liability as a substitute for approval-style oversight. The EU AI Act’s systemic-risk obligations have applied to frontier models since August 2025, and since August 2, 2025, the European Commission has had the power to demand evaluations and levy fines of up to 3% of a company’s global turnover for noncompliance. In other words, the jurisdiction Sacks might point to as proof that liability alone is sufficient is actually running both regimes, liability and pre-release oversight, on the same models at once. European lawyers have spent the second half of 2026 working through what that dual structure means in practice, from driver-assistance software to consumer chatbots, exactly the kind of overlap Sacks argues the US should avoid by picking liability only.
Sacks’ own tenure: 130 days, then out
Part of what gives Sacks’ comments weight is the position he no longer holds but clearly still shapes. He was appointed the White House’s first-ever AI and crypto czar in December 2024 and departed the formal role on March 26, 2026, after reaching the 130-day cap that governs special government employees, according to earlier Reuters and Block reporting on his exit. TIME reported he moved straight into a new post that same month, co-chairing the President’s Council of Advisors on Science and Technology, giving his September 13 remarks a level of institutional standing that a purely private commentator would lack.
He has not been the only departure from the administration’s AI apparatus this year. Chris Fall, director of the Center for AI Standards and Innovation, resigned after roughly three months on the job around July 2026, and Sriram Krishnan, a senior White House AI policy adviser, stepped down at the end of June 2026, though he has continued advising informally. That vacancy may not last: Reuters reported on September 19, 2026, that Trump was preparing to name a second AI czar and stand up a dedicated “AI force” inside the administration, a sign the policy gap Sacks has been filling informally is expected to get a formal successor. The churn has left Sacks as one of the more consistent voices connecting the White House’s public posture on frontier model access to the industry’s biggest labs, even without an official title.
The IPO backdrop nobody in this fight is ignoring
The timing of the safety debate is inseparable from Anthropic’s reported ambitions to go public at a valuation in the range of $2 trillion, a figure cited in coverage of the standoff by outlets including The News. If internal researchers at a company pursuing that kind of valuation are on record estimating a double-digit probability of AI-driven human extinction, that combination raises uncomfortable questions for regulators and prospective investors alike, and it helps explain why Sacks’ framing, that safety concerns are being used as leverage for favorable rules, has found an audience beyond the usual AI policy circles. Altman, for his part, has said OpenAI will not pursue an IPO in 2026, a decision that removes one point of direct comparison but leaves the underlying “pace the frontier” agreement between the two CEOs firmly in place.
Market and industry impact
For enterprise buyers and developers watching from outside the policy fight, the practical stakes are straightforward – and the shipping calendar has kept moving regardless of who wins the argument. MarkTechPost reported Google’s release of Gemini 3.8 Flash on September 2, 2026, the same day CNBC covered Meta’s launch of Muse Spark 1.3, while TechTimes logged both Anthropic’s Claude Opus 5.5 and OpenAI’s paired GPT-6 Sol and GPT-6 Luna models shipping on September 22, 2026, barely a week after Amodei and Altman’s public pledge to pace the frontier. If Sacks’ liability-first approach gains traction inside the administration, frontier labs would face fewer pre-release approval hurdles in the US but potentially larger legal exposure after a model ships and causes harm, a trade-off that favors well-capitalized incumbents who can absorb litigation risk over smaller challengers who cannot. If Amodei and Altman’s “pace the frontier” framing wins instead, the two companies could end up co-designing the regulatory guardrails that govern their own competitors, an outcome critics beyond Sacks have flagged as a competitive advantage dressed up as a safety measure.
Either path has consequences for the rest of the AI supply chain, from cloud providers hosting frontier models to enterprise software vendors building on top of Anthropic’s and OpenAI’s APIs. A liability-heavy regime raises the cost of insuring and indemnifying AI deployments across the board. An approval-heavy regime raises the cost and time of bringing new models to market, favoring companies that can already absorb months of review cycles.
Historical context: from AI czar to outside pressure
Sacks’ current role illustrates a pattern that has become common in Washington’s approach to fast-moving tech policy: officials cycle out of formal positions quickly, then continue to exert influence from advisory posts or private platforms. His 130-day term as AI and crypto czar was always going to be short by statute, but his subsequent co-chairmanship of the President’s Council of Advisors on Science and Technology has kept him inside the tent while freeing him to take public, adversarial positions against specific companies that a sitting czar might have had to soften. The administration backed that institutional continuity with money as well as titles: Axios reported that the White House launched a $100 million Innovation Council Action initiative in March 2026 to promote Trump’s AI agenda, the same month Sacks’ czar term lapsed. That dynamic has played out repeatedly in 2026, as CAISI’s Chris Fall and policy adviser Sriram Krishnan both left their roles within months of taking them, each citing different reasons but leaving a policy vacuum that voices like Sacks have partly filled.
What comes next: five predictions
- Anthropic and OpenAI are unlikely to formally withdraw their “pace the frontier” language, but expect both companies to reframe it publicly as a voluntary commitment rather than a request for new regulatory tools, to blunt Sacks’ cartel framing.
- METR’s independence will become a recurring talking point in Washington through the rest of 2026, with Sacks and allied voices continuing to question whether Anthropic-linked evaluators can serve as neutral arbiters of frontier-model risk.
- Congress is more likely to move toward a liability-focused framework than a pre-approval regime in the near term, given the administration’s stated preference and the lack of a unified industry position on approval-style rules.
- Anthropic’s reported IPO plans will draw closer scrutiny of internal safety statements like Evan Hubinger’s extinction-risk estimate, as underwriters and regulators weigh disclosure obligations tied to existential-risk claims made by the company’s own staff.
- Expect continued scrutiny of how AI-safety warnings spread on social platforms, following the questions Sacks raised about the Coxon resignation’s viral trajectory, as both critics and defenders of frontier labs increasingly treat virality itself as a data point in the policy fight.
Competitive landscape: how the major AI labs’ positions compare
| Entity | Public position on “pacing the frontier” | Regulatory preference | Key figure cited |
|---|---|---|---|
| Anthropic | Backs slowing model capability growth; internal alignment lead has flagged existential risk | Wants mechanisms allowing labs to pause/slow development | Evan Hubinger: greater than 10% extinction risk within a decade |
| OpenAI | Sam Altman publicly agreed with Amodei’s “pace the frontier” call | Aligned with Anthropic on voluntary pacing; ruled out a 2026 IPO | Sam Altman public endorsement of Amodei’s position |
| David Sacks / PCAST | Says labs can slow down unilaterally without new rules | Product liability and market discipline over approval regimes | Questions METR’s independence from Anthropic |
| European Union | Not applicable (regional bloc, not a lab) | Runs both strict product liability (from Dec. 9, 2026) and AI Act systemic-risk oversight (since Aug. 2025) concurrently | Up to 3% of global turnover in fines under the AI Act |
Timeline: key dates in the 2026 AI-czar standoff
| Date | Event |
|---|---|
| December 2024 | David Sacks appointed as the White House’s first AI and crypto czar |
| February 2025 | European Commission withdraws its proposed AI Liability Directive |
| August 2, 2025 | EU AI Act systemic-risk enforcement powers, including turnover-based fines, take effect |
| March 26, 2026 | Sacks’ 130-day term as AI and crypto czar expires; he departs the formal role |
| June 2026 | Sacks accuses Anthropic of a “regulatory capture scheme” over its AI-pause proposal |
| Late June 2026 | Sriram Krishnan steps down as senior White House AI policy adviser |
| July 2026 | Sacks criticizes Anthropic’s stance on open-weight models; Chris Fall resigns as CAISI director |
| September 8, 2026 | Anthropic pretraining researcher Jacob Coxon resigns, warns industry is “gambling with people’s lives” |
| September 13, 2026 | Sacks tells Anthropic and OpenAI to slow down on their own and stop asking for antitrust waivers or approval regimes; The Next Web reports the exchange |
| December 9, 2026 | EU’s strict product-liability rules for software and AI systems take effect |
What this means for developers and enterprise AI buyers
None of this changes what ships in production today, but it shapes the regulatory environment developers will be building against over the next 12 to 24 months. Teams deploying Anthropic or OpenAI models in production should watch for two concrete signals: whether either company voluntarily delays a flagship release citing safety review, and whether Congress or the administration moves toward codifying a liability standard resembling the EU’s December 9 rules. A liability-first US framework would likely mean more detailed model cards, more explicit terms-of-service disclaimers around cybersecurity misuse, and potentially higher API pricing to cover insurance costs, rather than new pre-release certification requirements that could slow model availability.
Frequently asked questions
Who is David Sacks?
David Sacks is a venture capitalist who served as the White House’s first AI and crypto czar from December 2024 until March 26, 2026, when his 130-day term as a special government employee reached its limit, as Reuters reported. He now co-chairs the President’s Council of Advisors on Science and Technology, a post TIME reported he took up in March 2026 immediately after leaving the czar role.
What exactly did Sacks tell Anthropic and OpenAI?
He said the companies could go ahead and slow down frontier model development on their own, since they already sit at the top of the market, but should stop asking for an antitrust waiver or a new government approval regime in exchange, calling that kind of trade-off effectively “blackmail.”
Why are Anthropic and OpenAI talking about “pacing the frontier” now?
Anthropic CEO Dario Amodei publicly argued the industry needs to slow the pace of capability improvements, and OpenAI CEO Sam Altman agreed that OpenAI would do the same, a rare public alignment between two rival labs.
Who is Jacob Coxon and why does his resignation matter?
Coxon was a pretraining researcher at Anthropic who resigned on September 8, 2026, saying publicly that the AI industry was “gambling with people’s lives.” His departure intensified public debate over AI safety just before Sacks’ comments.
What did Anthropic’s own safety team say about existential risk?
Evan Hubinger, who leads Anthropic’s Alignment Science team, has publicly estimated the probability of AI causing human extinction within the next decade at more than 10%.
Is Anthropic actually going public?
Anthropic has been reported as pursuing a public listing at a valuation in the range of $2 trillion, according to coverage tied to this week’s safety debate, though no formal IPO date has been confirmed.
How does the EU handle AI liability differently than the US?
The EU withdrew a dedicated AI Liability Directive in February 2025 and instead extended existing strict product liability rules to cover standalone software and AI systems starting December 9, 2026, while separately keeping AI Act approval-style oversight in force since August 2025, running both systems at once.
What is METR and why did Sacks question it?
METR is a third-party organization that evaluates frontier AI models for dangerous capabilities, and Anthropic uses it as part of its safety testing. Sacks questioned METR’s independence, arguing it is too closely tied to Anthropic’s investors and staff to serve as a neutral evaluator.
