Anthropic’s IPO tests its safety stance

The Frontier

Science & tech breakthroughs in AI models, hardware, and computing that define what's possible

Anthropic Pursues an IPO Despite Its AI Safety Warnings. Anthropic is pursuing an initial public offering while projecting it could reach $100 billion in annualized revenue during 2026, according to the New York Times. The prospective listing would bring a leading frontier-model developer into public markets even as CEO Dario Amodei has advocated slowing development of some advanced AI systems.

Anthropic’s $100B annualized-revenue projection makes its safety stance a credibility test, not a brake: more than 25 times Facebook’s $3.7B IPO revenue. The test by 2026 is whether its warnings still constrain advanced-model development as it prepares to list.

New York Times

Anthropic Is Operating a Lab That Conducts Biology Experiments. Anthropic confirmed that it operates a Bay Area wet biology laboratory where its AI models can support physical experiments. The company said the facility focuses on fundamental biology and follows its April acquisition of AI biotech company Coefficient Bio.

Anthropic’s Bay Area lab is a bid to own experimental data rather than depend on third-party biology data. The proof point is a public example linking a lab experiment to a model capability unavailable from third-party data.

TechCrunch

FAA Prepares $875M AI Tool for Air-Traffic Congestion. The FAA is preparing to deploy SMART, an $875 million AI system designed to forecast air-traffic flows and identify operational conflicts. The initial rollout is expected to cover Washington-area airports before a planned expansion across the FAA's national airspace system.

FAA’s $875M SMART rollout at Washington-area airports will test whether AI forecasts can identify operational conflicts before the system expands across the national airspace system.

Ars Technica

Model Developments

Latest AI model releases, capabilities, and technical breakthroughs shaping the competitive landscape

Anthropic Taps Accenture’s Faculty for Embedded AI Model Evaluation. Anthropic has partnered with Accenture's Faculty AI unit to conduct embedded evaluation of frontier models, including red-teaming, alignment assessments, and safeguard testing. The companies said they expect to invest at least $1 billion each over five years to build evaluation capacity, with details of the operating model still being developed.

Anthropic is moving Accenture Faculty beyond AI integration into frontier-model evaluation. The companies expect to invest at least $1B each over five years; the test is whether the operating model embeds Faculty in red-teaming, alignment and safeguard testing.

Unite.ai

Vals, Backed by Andreessen Horowitz, Wants to Set the AI Benchmarking Standard. AI benchmarking startup Vals raised a $40 million Series A led by Andreessen Horowitz after a seed round led by 8VC and Bloomberg Beta. Founded in 2024, Vals is developing evaluation methods intended to measure modern AI systems more reliably as companies increasingly optimize against established benchmarks.

Vals’ $40 million Series A makes independent AI evaluation an infrastructure bet; the test is whether its methods measure modern AI systems more reliably than established benchmarks.

TechCrunch

Capital & Control

Tracking the flow of capital and influence that shapes the tech landscape

FCC Lets Paramount Sell a 49.5% Equity Stake to Saudi Arabia, UAE, and Qatar. The Federal Communications Commission approved Paramount Skydance's request to raise indirect foreign ownership to 49.5% through investments from Saudi Arabian, UAE, and Qatari sovereign wealth funds. The approval applies to Paramount's broadcast-station licenses and comes as the company pursues its $111 billion acquisition of Warner Bros. Discovery.

Paramount’s FCC approval creates a financing channel for its Warner Bros. Discovery bid, but no sovereign funding is secured yet. Watch for Saudi, UAE, or Qatari funds named in any acquisition financing announcement.

Ars Technica

Infrastructure & Power

The physical backbone of AI: compute, data centers, supply chains, and the geopolitics of energy

Spanberger Signs Data Center Accountability Order and Creates AI Task Force. Virginia Governor Abigail Spanberger signed Executive Order 22, creating a Data Center Accountability Framework that directs state action on data-center transparency, environmental review, and energy costs. The order also establishes a rapid-response AI task force, while proposed legislation is planned for the 2027 General Assembly session.

Spanberger tightens oversight, but NOVA data centers already draw 2 GW and each new AI campus wants another 500 MW, enough for all of Richmond. Transmission takes 3-7 years, buildings just 2. Net effect: projects pause or leave Virginia unless utilities lock in extra capacity by 2027.

Unite.ai

Digital Defense

Threats and defenses shaping the global security landscape

AI Hallucination Nearly Led to U.S. Military Action. CNN reported that a U.S. Special Operations Command analyst used a chatbot to assess intelligence on a Chinese vessel, producing a false conclusion that it carried nuclear-program components. U.S. forces reportedly prepared to intercept and board the ship before officials found that the AI had misidentified the cargo.

CNN’s report makes independent verification a prerequisite for chatbot-derived intelligence: a false cargo assessment reportedly put U.S. forces on standby to board a Chinese vessel. Watch for DoD validation gates by Q1 2027.

Ars Technica

The Performance Edge

Peak performance science for sustained execution in high-stakes environments

Ozempic Can Change Your Body. Can It Change Your Life? University of Chicago researchers found limited evidence that GLP-1 medicines such as Ozempic produce large improvements in mental health, employment, marriage, or overall well-being despite established benefits for blood sugar control, weight loss, and cardiovascular health. The research separates physical-health outcomes from broader social and psychological measures often associated with the drugs’ rapid adoption.

1: University of Chicago’s findings make GLP-1s a health intervention, not a proven life-reset. Claims of better jobs, marriages or mental health lack support; the test is whether multi-year follow-up finds durable gains beyond physical markers.

Science Daily

Crypto & Digital Assets

Bitcoin, Ethereum, DeFi protocols, and digital asset market dynamics

North Korean WaterPlum Hackers Infected 30,000 Devices Worldwide. A joint law-enforcement advisory said the North Korean-linked WaterPlum group compromised at least 30,000 devices globally between December 2025 and July 2026. Authorities said the campaign transferred more than $10.7 million in stolen cryptocurrency to North Korea.

WaterPlum’s 30,000-device campaign and $10.7M transfer do not justify endpoint, crypto-custody, or cyber-insurance repricing. Watch for a joint advisory by July 2027 documenting another campaign at similar scale.

Bleepingcomputer

Market Pulse

Daily market data: indices, sectors, top movers, and volatility

Markets Closed for the Weekend. U.S. equity markets are closed for the weekend. The next market update arrives Monday at 6:30 a.m. Pacific, when the NYSE opening bell rings.

NYSE’s Monday 6:30 a.m. Pacific opening bell is the next scheduled U.S. equity price signal; no new U.S. equity pricing is available before then. Watch for the first post-weekend prices at Monday’s open.

Muninn