AI News Flash · Daily Brief

Anthropic surpasses OpenAI in valuation after closing $65B Series H round.

By and · Editorial team, AI News Flash

Platforms

Anthropic surpasses OpenAI in valuation after closing $65B Series H round.

Anthropic closed a $65 billion Series H on May 28, led by Altimeter Capital and Sequoia, lifting its valuation to $965 billion and placing it ahead of OpenAI among private AI companies. The fundraise follows strong revenue momentum: the company's annualized run rate crossed $47 billion earlier in May, with Claude Code estimated to hold roughly 54% of the enterprise AI coding market. The deal cements Anthropic's position as a dominant commercial force in the frontier AI space and sets the stage for a reported October 2026 IPO, a milestone that would mark one of the largest technology listings in recent history.

Why it matters: Enterprises and investors now face a two-horse frontier AI race with a near-trillion-dollar private incumbent ahead of a major IPO.

Microsoft teases proprietary coding model ahead of Build 2026

Microsoft announced on May 28 that it will reveal new proprietary AI models at its Build 2026 conference on June 2 and 3, headlined by a coding model pitched as a cost-effective alternative to Claude Code and OpenAI Codex. The move is made possible in part by a renegotiated partnership with OpenAI that removed Microsoft's right of first refusal as OpenAI's sole compute provider, giving the company cleaner competitive footing to develop and distribute its own models. If the coding model delivers on its cost positioning, it could directly pressure Anthropic's estimated 54% share of the enterprise AI coding market and reduce Microsoft's dependence on third-party frontier labs.

Why it matters: Enterprises evaluating AI coding tools will soon have a lower-cost, Microsoft-native option competing directly against Anthropic and OpenAI offerings.

Capabilities

Claude Mythos Preview scores 64.7% on HLE, opening a 12.6-point lead over GPT-5.4 Pro.

Claude Mythos Preview now leads the Humanity's Last Exam leaderboard at 64.7% with tools enabled, compared to 52.1% for GPT-5.4 Pro, a gap of 12.6 percentage points on a benchmark engineered to remain difficult as models improve. Without tools, Mythos scores 56.8%, up from Opus 4.6's 40% on the same benchmark, representing a 42% relative gain. The results are self-reported by Anthropic and have not yet been independently verified through Scale AI's official leaderboard. Access to Mythos remains limited to Glasswing security partners, meaning broader performance validation and real-world enterprise assessment are still pending.

Why it matters: AI developers and enterprises should note that the leading benchmark score is self-reported and unverified, making independent validation a critical next step.

Technology & Research

NVIDIA's open-source Ising models cut quantum error-correction time by up to 2.5x.

NVIDIA has released Ising, the first open-source family of AI models purpose-built for quantum processor calibration and error correction, two of the primary bottlenecks preventing quantum computers from reaching practical utility. The models deliver up to 2.5x faster and 3x more accurate error-correction decoding compared to traditional approaches, integrate with NVIDIA's CUDA-Q platform and NVQLink QPU-GPU interconnect, and can run locally to protect proprietary research data. Early institutional adopters include Harvard and Fermi National Accelerator Laboratory. By open-sourcing the models, NVIDIA is positioning its hardware and software stack as essential infrastructure for the quantum computing research community at a critical stage of the field's development.

Why it matters: Quantum computing researchers and hardware developers now have an open, faster error-correction baseline that could accelerate the path to practical quantum utility.

Regulation & Policy

OpenAI publishes Frontier Governance Framework mapping to EU AI Act

OpenAI released its Frontier Governance Framework on May 28, a formal document that maps the company's internal safety and governance practices to two specific regulatory instruments: the EU AI Act's Code of Practice for General-Purpose AI and California's Transparency in Frontier AI Act. The publication makes OpenAI the first major US frontier AI laboratory to explicitly cross-reference its internal governance to the EU's GPAI tier obligations in a public document. The move signals a deliberate strategy to demonstrate compliance readiness ahead of the full enforcement deadlines for both the EU and California frameworks, and sets a new transparency benchmark that regulators and competing labs are likely to assess closely.

Why it matters: Regulators and competing AI labs now have a concrete governance template to measure against as EU AI Act and California enforcement deadlines approach.

Federal court extends xAI's stay to Colorado's revised AI discrimination law SB 26-189.

A federal court has explicitly extended the stay in x.AI LLC v. Weiser to cover SB 26-189, the revised AI algorithmic-discrimination statute Colorado Governor Polis signed on May 14, 2026, as a narrower replacement for the original SB 24-205. Both laws are now unenforceable while the constitutional challenge continues. The case is notable as the first instance of a federal court suspending a state AI law at the request of a private AI company, with the Department of Justice intervening in support of xAI's position. The precedent could complicate enforcement of similar AI regulation in other states where comparable algorithmic-discrimination or transparency bills are advancing.

Why it matters: AI companies operating in multiple states now have a federal precedent supporting court challenges that can halt state AI regulations before they take effect.