Big Tech Unites for Open Source AI—and Against Anthropic

July 28, 2026

ai-daily-brief-podcast

Overview

This episode of the AI Daily Brief (recorded July 28, 2026) covers two major stories: NVIDIA’s investment in Ilya Sutskever’s Safe Superintelligence (SSI), and the emergence of a broad Big Tech coalition publicly supporting open-weight AI models — with Anthropic conspicuously absent. The host (Nathaniel Whittemore, implied by the show’s known format) frames the open-vs-closed AI debate as having reached a new political and commercial inflection point. No guest speakers are featured.

Source video: URL not provided in submission.

Prerequisites

Basic familiarity with the open-source vs. proprietary software debate and how it maps onto AI model development
Understanding of what open-weight models are (models whose weights are publicly downloadable and modifiable)
Awareness of key AI labs: OpenAI, Anthropic, Google DeepMind, Meta, NVIDIA, Safe Superintelligence (SSI)
Background on U.S.–China chip export controls (ASML, TSMC, NVIDIA export restrictions)
Familiarity with the term model distillation (using one model’s outputs to train/improve another)
General awareness of U.S. federal AI policy debates (executive orders, entity lists, Commerce/Treasury actions)

Main Points

1. NVIDIA Invests in Safe Superintelligence (SSI)

NVIDIA made a “substantial” but undisclosed investment in SSI, the lab founded by former OpenAI Chief Scientist Ilya Sutskever after he departed following the late 2023 OpenAI leadership crisis.
SSI will receive access to NVIDIA’s next-generation Vera Rubin chips, 10x-ing its compute resources over the next 12 months; SSI had previously relied on Google TPUs.
NVIDIA stated it gained “rare access” to SSI’s technology before investing; Jensen Huang credited Ilya with pioneering “fundamental breakthroughs at the foundation of modern AI.”
SSI’s stated mission from inception has been a “straight-shot” to safe superintelligence, deliberately avoiding short-term commercial distractions — a strategic contrast to competitors like Mira Murati’s Thinking Machines Lab, which has pivoted toward enterprise fine-tuning services.
The announcement signals SSI has reached a stage where its research is “worth scaling,” according to the company’s own statement.

2. NVIDIA’s Circular Financing Concerns and China Chip Race

NVIDIA stock fell ~4% amid scrutiny of $3.25 trillion in deals perceived as “circular”: a $500B+ partnership with SK Group and a $250B backstop for an OpenAI data center.
On the geopolitical front, China has rapidly reduced its dependence on U.S. chips: from 90% reliance in 2021 to 60% in 2025, with Morgan Stanley projecting a potential drop to 25% by ~2030.
Huawei held a closed-door demo of next-gen chips showing dramatic progress; Chinese Communist Party officials reportedly warned major AI users that resisting domestic chips amounted to “treason.”
A Shanghai-based company has begun mass-producing deep ultraviolet (DUV) lithography machines — previously unavailable to China under export controls — marking a significant milestone in China’s domestic chip supply chain.
Bernstein forecasts China’s chip manufacturing throughput will double annually for three years, with Huawei alone expecting to ship 1.5 million AI chips in the current year. Despite this, Bernstein still expects China to lag the U.S. through at least 2030.

3. Apple vs. Micron: Memory Chip Lobbying Battle

Apple is lobbying the White House to allow purchases of memory chips from Chinese manufacturer CXMT, arguing that memory prices are unsustainable and threaten consumer pricing (invoking the specter of “$5,000 iPhones”).
Micron, the only sizable U.S. memory chip manufacturer, is counter-lobbying to keep restrictions in place, warning that Chinese access to the U.S. market could “decimate” domestic production — drawing parallels to the decline of U.S. steel.
Technically, no White House approval is required; the only current restriction is the Pentagon entity list, applying only to the military supply chain. However, Apple is reluctant to act unilaterally during a politically tense period.
The Trump administration faces a direct policy tension: combating inflation (worsened by high memory prices) vs. promoting domestic chip manufacturing. No clear resolution has been announced.

4. The Political Context: Washington Moves Toward an Open-Source AI Ban

Concerns about Chinese open-source models have escalated in Washington, driven by the release of GLM 5.2 and especially Kimi K3 (a 2.8-trillion-parameter open-weight model that outperformed leading U.S. models on at least one benchmark).
Treasury Secretary Scott Bessent and senior tech official Michael Kratzios publicly foreshadowed potential sanctions or entity list designations against Chinese firms conducting what they termed “covert, industrial-scale distillation attacks” on American IP.
Commerce Secretary Howard Lutnick was reportedly the lone senior administration official still advocating for U.S. open-source competitiveness as the preferred response, and was taking meetings with AI labs ahead of policy decisions.
The cybersecurity angle shifted the discourse: it was later revealed that Kimi K3 significantly lagged U.S. models on key cybersecurity benchmarks, undercutting some of the urgency around a ban.

5. Big Tech’s Open Letter in Support of Open-Weight Models

A broad coalition of Big Tech companies released an open letter arguing that open-weight AI models are essential to U.S. technological leadership, competition, and economic diffusion of AI benefits.
The letter drew historical parallels to the open-source software movement of the 1980s, arguing that open foundations enabled generations of American innovation.
Key signatories and public endorsers included: Jensen Huang (NVIDIA), Sundar Pichai (Google), Satya Nadella (Microsoft), Mark Zuckerberg (Meta), and Elon Musk (xAI); dozens of other companies also signed.
The letter explicitly addressed security concerns, arguing that open models broaden defensive cybersecurity capability and that restricting them would create a single point of failure, undermining safety.
On model distillation specifically, the letter drew a distinction between lawful distillation (a standard technique for model improvement with a long tradition in ML) and unlawful extraction of closed model weights — arguing only the latter warrants legal remedy.

6. OpenAI’s Ambiguous Position and Eventual Endorsement

OpenAI was initially not a signatory to the open letter, which prompted significant public criticism.
Sam Altman responded positively on social media (“I want the U.S. to win in AI both in open source and proprietary models”) and OpenAI was subsequently added to the list of signatories.
Altman has separately expressed concern about AI “authoritarianism” and the dangers of a small number of companies controlling AI — views somewhat aligned with the letter’s thesis.

7. Anthropic’s Refusal to Sign — and the Resulting Debate

Anthropic was the sole major U.S. AI company not to sign the letter, and this became the central point of industry discourse.
Anthropic CEO Dario Amodei has publicly warned about risks from open models, specifically citing concern that “Mythos-class cyber abilities available for anyone to download” would represent a serious threat. The company is reportedly working with the government on countering distillation attacks.
Former AI czar David Sacks characterized the situation as “the entire tech industry, save for Anthropic, in favor of open source AI,” and accused Anthropic of pursuing a regulatory strategy to “kneecap” open models.
Defenders of Anthropic’s stance argued: (a) they are acting on genuine principle consistently over time; (b) they could have signed and lobbied against it privately but chose not to; and (c) if an open-weight model is ever involved in a serious safety or security incident, Anthropic’s position will be vindicated.
A member of Anthropic’s Technical Staff (Julian Schritweiser) posted sarcastic tweets suggesting Jensen Huang and Satya Nadella should open-source CUDA drivers, Windows, and MS Office if they truly believe in open source — generating significant backlash, including from investors and commentators who argued it revealed Anthropic’s commercial motivations.
The internal lab perspective (articulated via an OpenAI employee’s tweet) holds that “there is no way to hold a consistent belief set where you’re AGI-pilled and pro-open source” — implying those who deeply believe in transformative AGI risk are logically compelled to oppose open-weight models.

Key Concepts

Open-weight models: AI models whose trained parameters (weights) are publicly released, allowing anyone to download, inspect, modify, and run them on their own infrastructure — distinct from open-source code alone.
Model distillation: A technique in which the outputs of a larger or more capable model are used to train or improve a smaller or different model; widely used in ML research and development.
Safe Superintelligence (SSI): A research-focused AI lab founded by Ilya Sutskever with the singular stated mission of building safe superintelligence, deliberately eschewing short-term commercial products.
Vera Rubin chips: NVIDIA’s next-generation AI accelerator platform, succeeding the Hopper and Blackwell architectures.
Kimi K3: A 2.8-trillion-parameter open-weight model released by Chinese AI lab Moonshot AI; the largest open-source model released at the time of this episode, notable for outperforming some U.S. models on specific coding benchmarks.
Deep Ultraviolet (DUV) lithography: A chip manufacturing technology previously unavailable to China under U.S. export controls; China has now begun domestic mass production of DUV machines.
ASML: Dutch company and monopoly supplier of advanced chip fabrication equipment (including EUV lithography); access to its most advanced machines has been denied to China.
Entity list: A U.S. Commerce/Pentagon list of foreign entities restricted from receiving certain U.S. exports or doing business with U.S. companies in specified ways.
Circular financing: A concern raised about interconnected investment flows in the AI industry (e.g., chip companies investing in AI labs that then spend that capital buying chips), which critics argue inflates apparent market activity.
Little Tech Association: A U.S. trade group representing startups (including Y Combinator members) that lobbied against a ban on Chinese open-source AI models, arguing it would cause a “mass extinction event” for U.S. startups building on those models.
AGI-pilled: Colloquial term for the belief that artificial general intelligence (and beyond, superintelligence) poses near-term transformative or existential-level risks, informing a cautious stance on model openness.

Summary

The episode documents a pivotal moment in U.S. AI policy: a broad coalition of Big Tech companies — including NVIDIA, Google, Microsoft, Meta, and eventually OpenAI — publicly aligned behind open-weight AI models through a joint open letter, framing openness as essential to American technological competitiveness, innovation, and even AI safety. The coalition formed in direct response to Washington policymakers’ growing inclination to restrict Chinese open-source models (driven by concern over distillation attacks and model releases like Kimi K3), and was galvanized by fears that such restrictions would entrench the market dominance of closed frontier labs. Anthropic’s conspicuous refusal to sign the letter crystallized a genuine philosophical divide: the company’s leadership genuinely believes that powerful open-weight models represent a serious safety and security risk, while critics contend that Anthropic is using safety rhetoric to protect its commercial position through regulatory capture. The episode frames this not as a simple good-vs-evil conflict but as a collision of legitimately competing worldviews — one holding that diffusion and democratization of AI is inherently safer, the other holding that the proximity to AGI makes openness genuinely dangerous — with enormous policy, competitive, and safety consequences riding on which view prevails in Washington.