Anthropic reported that Claude produced the first complete, computer-checked formalization of Fermat’s Last Theorem in the Lean theorem prover. Working largely autonomously over 11 days, multi-agent teams wrote roughly 13 million lines of Lean and proved about 29,500 intermediate theorems, relying only on Lean’s standard axioms.

OpenAI separately expanded its cyber-defense program with Daybreak for Frontline Defenders, a $1 billion commitment of subsidized Daybreak access, training, and support aimed at operators of essential services. GitHub also made GPT-6 Astra generally available in Copilot for Pro+, Max, Business, and Enterprise users across major IDEs and surfaces.

Together the releases frame a weekend in which machine-checked mathematics, governed cyber tools for frontline defenders, and frontier coding models moved from research and limited access into broader product channels.

Humanoid Press is thrilled to be the official media partner of the AI Infra Summit 2026. The event will bring together leaders from hyperscalers, chipmakers, and enterprise IT to discuss the critical infrastructure needed to support the explosive growth of AI.

Keynote recordings and full agenda details are now available. Stay tuned for exclusive coverage from the Summit.

Humanoid Press looks forward to live coverage, keynote recaps, and in-depth analysis that will help our readers understand how the AI infrastructure stack is evolving. We expect an event of this scale to deliver transformative insights into the systems that will define tomorrow’s digital economy.

With coverage from the event, our readers will gain a clear view of the infrastructure that will power the next wave of AI innovation and enterprise adoption.

Anthropic reported that Claude produced the first complete, computer-checked formalization of Fermat’s Last Theorem in the Lean theorem prover. Working largely autonomously over 11 days, multi-agent teams wrote roughly 13 million lines of Lean and proved about 29,500 intermediate theorems, relying only on Lean’s standard axioms.

OpenAI separately expanded its cyber-defense program with Daybreak for Frontline Defenders, a $1 billion commitment of subsidized Daybreak access, training, and support aimed at operators of essential services. GitHub also made GPT-6 Astra generally available in Copilot for Pro+, Max, Business, and Enterprise users across major IDEs and surfaces.

Together the releases frame a weekend in which machine-checked mathematics, governed cyber tools for frontline defenders, and frontier coding models moved from research and limited access into broader product channels.

Google made Lyria 3.5, its highest-fidelity music generation model, available in the Gemini app and the Gemini API. The release adds more expressive vocals, richer arrangements, genre and vocal/instrumental controls, creative templates, and a choice of short or longer tracks for everyday and professional use.

Microsoft began rolling out GPT-6 Astra through the Microsoft Foundry Limited Access Program, with expansion to participating enterprise customers over the coming days and Standard Global and U.S. Data Zone deployment options. Separately, Google’s agentic video understanding path for Gemini models continues to emphasize lower token cost and higher accuracy on long-form video analysis.

Together the moves underline a week in which generative media, enterprise frontier distribution, and efficiency-focused multimodal tooling advanced in parallel across consumer apps and cloud platforms.

NVIDIA announced it has agreed to acquire Hugging Face for $12,930,300,000. CEO Jensen Huang said the deal will scale the platform’s infrastructure and expand access for developers and institutions while keeping the hub’s open character intact. Closing is expected subject to regulatory review.

OpenAI began rolling out GPT-6 Astra, its most capable model to date, first to a limited set of organizations with broader ChatGPT and API access planned over the coming days. Astra is priced at $10 / $50 per million input/output tokens and is the first model OpenAI designates at the Critical cyber threshold under its Preparedness Framework.

Separately, MBZUAI’s Institute of Foundation Models released K2 Horizon — six Apache 2.0 models from 0.9B to 375B-A23B — with weights, training code, checkpoints, and data recipes, framed as one of the most comprehensive open releases to date.

Google DeepMind released Gemini 3.8 Flash, its most intelligent workhorse Flash model yet, with gains on software engineering, agentic tasks, and multi-step reasoning at the same introductory price as 3.7 Flash. A parallel Gemini 3.8 Flash Cyber variant is available to trusted defenders through the new Fairwind Program for vulnerability discovery and automated patching.

Meta Superintelligence Labs shipped Muse Spark 1.3 in Muse Code and the Meta Model API, citing fewer tool calls and lower token overhead on coding workflows versus 1.2. OpenAI expanded ChatGPT for Healthcare with a read-oriented Epic EHR integration and a Healthcare Public Data plugin for structured access to official sources such as PubMed and CMS Coverage.

Together the launches underscore a week defined by faster workhorse models, specialized cyber tiers, and tighter wiring of frontier systems into enterprise and clinical workflows.

Anthropic released Claude Fable 5.1 for broad use and Claude Mythos 5.1 for trusted-access programs in cybersecurity and life sciences. The pair share one underlying model with different safeguard layers; Fable 5.1 is positioned for coding and knowledge work at an estimated 25% lower cost for typical token-billed workloads than Fable 5.

Separately, OpenAI said further evaluations show Astra meets the Critical cybersecurity capability threshold under its Preparedness Framework, including a perfect score on ExploitBench, and outlined stronger gates before any release of the most advanced cyber capabilities. DeepSeek published open weights for V4-Flash-Vision-Exp, its first experimental multimodal V4 model, under an MIT license on Hugging Face.

Anthropic also previewed Enterprise Frontier Safeguards for customer-controlled storage with misuse detection, while Fable 5.1 remains available across the Claude Platform and major cloud marketplaces.

The AI Infra Summit has posted its keynote recordings for on-demand viewing. Hyperscaler sessions from AWS, Meta, and Google cover how those operators are rebuilding the stack for AI at scale. Separate panels bring enterprise voices from CVS Health, Pfizer, and Lambda, and a data-center track with Microsoft, Cologix, NVIDIA, and Schneider Electric.

Next-generation compute keynotes include Lightmatter on photonic AI infrastructure and Cadence on systems design — the same themes that will carry into the live program. Click the official mark in the panel to open the keynote archive. Resource-center reports and interviews sit alongside the video set for teams tracking power, networking, and deployment.

The 2026 conference runs September 15–17 at the Santa Clara Convention Center, with main-stage, data-and-models, compute, data-movement, and AI data-center tracks. Agenda and speaker lists are live for those planning the full week on site.

Sony Music Publishing, Warner Chappell and allied publishers filed suit in the Northern District of California against Anthropic, CEO Dario Amodei, and co-founder Benjamin Mann, alleging illegal torrenting, scraping, and downloading of copyrighted compositions to train Claude. The complaint seeks statutory damages and an accounting of training data.

On the open-source side, DeepSeek’s Harness (dsh) remains the week’s standout agent runtime — MIT-licensed, everything-is-a-plugin, and model-agnostic. Meta’s Muse Glimmer, a 30B open-weight agentic model under Apache 2.0, continues to define the local-device alternative for always-on tool use without cloud dependency.

Together the filings and releases frame a weekend in which copyright risk and composable agent infrastructure moved in parallel across closed labs and open stacks. Anthropic has said it will defend the case; the publishers seek jury trial and up to statutory maximums per work, keeping training-data accountability at the center of the dispute.

OpenAI notified SpaceX that it will wind down the contract supplying models to Cursor, with a proposed shutoff on November 12, 2026. The company said it cannot be confident SpaceX will stay within terms of service, citing past violations by Musk-owned companies including xAI, and accountability needs around the upcoming Astra model.

Separately, Anthropic reported that automated alignment researchers built on Claude mitigated all ten tested alignment failure categories without degrading capabilities, generalizing to held-out benchmarks and larger models. Tencent released Hy4 preview open source: 770B total parameters, 49B active, 1M context, aimed at coding, office work, and research.

The week underscores how change-of-control clauses, self-improving safety loops, and open-weight scale are reshaping both commercial access and the research stack for frontier systems.