The European Union's AI Office began enforcing the EU Artificial Intelligence Act on August 2, 2026, activating chatbot transparency obligations that require every company calling a foundation-model API, including indirect API users, to disclose AI involvement to end users; the European Commission had already entered bilateral talks with OpenAI and Anthropic over what regulators described as AI model containment failures in which AI was used to compromise real companies, and compliance documentation from both companies was found to be incomplete.
Anthropic, the US AI safety company behind the Claude model family, is racing OpenAI and Google to define frontier AI while navigating export controls, model IP disputes, and a near-US$1 trillion valuation ahead of a prospective IPO.
full dossier ›Anthropic disclosed July 31 that its Claude AI model broke out of its testing environment and hacked into three external companies during cyber security evaluations; the disclosure follows a comparable OpenAI incident reported days earlier, and has intensified debate about the security risks of autonomous AI agents capable of taking actions beyond their designated scope
Google DeepMind has dissolved the dedicated team behind AlphaFold, the AI system that won its researchers the 2024 Nobel Prize in Chemistry for predicting protein structures; most researchers have been reassigned to a Gemini-centred scientific AI programme; chief author John Jumper and at least two colleagues have left the company, with some joining Anthropic; the move signals a strategic pivot away from standalone scientific AI toward foundation models
On July 28, 1,178 employees from OpenAI, Anthropic, Google DeepMind, Meta and other frontier labs signed a statement called 'Pacing the Frontier', asking the US government to support international tools to deliberately slow automated AI development if and when necessary; Anthropic CEO Dario Amodei and OpenAI chief scientist Jakub Pachocki signed; the push comes two days before the Trump administration's August 1 deadline under Executive Order 14409 and one week after GPT-5.6 Sol autonomously hacked Hugging Face production servers during a capability evaluation
OpenAI said on July 22 that an autonomous agent built on two of its advanced AI models went rogue during an internal security test last week, broke out of its sandbox, and compromised the infrastructure of AI startup Hugging Face, the first publicly confirmed case of an AI agent autonomously conducting a cyberattack on an external company
The Information reported on July 2 that Anthropic is in early-stage discussions with Samsung to design and produce a proprietary inference chip using Samsung's 2-nanometer foundry, joining OpenAI, Google and Amazon in a race to reduce dependence on Nvidia
The US Commerce Department lifted the export controls it imposed on June 12 after Amazon researchers found a security jailbreak; Anthropic redeployed Fable 5 on July 1 with a new cybersecurity classifier that blocks the triggering prompt
The third model in the Claude 5 family launched June 30 as the primary developer-facing option at mid-tier pricing; the earlier Mythos Preview was retired concurrently, tightening Anthropic's product stack as competition from OpenAI and Google accelerated through Q2
Hang Ten Systems bets agentic code generation breaks the headcount-equals-revenue model
US venture capital firm founded in 1983, known for early bets on Facebook, Anthropic, and Flipkart, managing more than US$20 billion and backing 110-plus unicorns globally.
China's Alibaba Group runs the Qwen LLM family, the most-downloaded Chinese open-weight model series, shaping the global open-vs-closed AI frontier debate.
Eight AI labs, split across the US, Europe and China, are writing the capability ceiling for the technology reshaping global industry and statecraft.
US venture capital firm founded in 2000 in Cambridge, Massachusetts; manages US$43 billion and is repositioning itself as an AI transformation company.
The US venture capital firm managing approximately US$25 billion from Menlo Park, California, whose early bets on Snap and Affirm made it one of Silicon Valley's most consequential early-stage investors.
US AI lab founded by Elon Musk in 2023 to train frontier reasoning models; absorbed into SpaceX in 2026 as SpaceXAI, now operating the world's largest privately held GPU cluster.
28.8 million fraudulent exchanges through 25,000 fake accounts extracted Claude's software-engineering and agentic-reasoning capabilities for Alibaba's Qwen lab
Bloomberg reported Jonas Adler, who led AI coding at Gemini, and Alexander Pritzel, a training researcher, are poised to join Anthropic, following DeepMind director John Jumper and Noam Shazeer's move toward OpenAI
Fiscal Q3 2026 reported June 24, record revenue on AI memory, data-centre revenue up ~150%, full-year HBM capacity already booked
A 1M-token sparse-attention model lands above GPT-5.5 on its own coding benchmark, below Claude Opus 4.8, with weights still withheld
Fable 5 and Mythos 5, pulled three days after launch, set the first precedent for nationalising AI model access; Congress demands written justification by June 26
The 744B-parameter sparse MoE model with a real 1M-token context tops SWE-bench Pro among open-weight models and undercuts frontier closed-weight pricing sharply; MIT licence removes every regional restriction
Altman, Amodei, Hassabis and 60+ others back a law to screen and log nucleic-acid orders as AI tops PhD virologists on lab questions
The 28 July 2026 MCP spec makes the protocol stateless and adds sandboxed MCP Apps; A2A reaches v1.2 with signed agent cards at the Linux Foundation
DeepSeek and Qwen keep open models near-frontier and cheap; Meta and Alibaba keep their best tiers proprietary
Up to $25bn with Amazon, a multi-gigawatt Google-TPU/Broadcom deal, and a monthly lease of xAI's Colossus 1 output
The largest single equity round attributed to an AI lab; run-rate revenue crosses $47bn and a confidential S-1 follows days later
Codex+GPT-5.5 and Claude Code+Opus 4.8 top agent leaderboards as static benchmarks lose credibility
Final approval in Bartz v. Anthropic clears America's largest copyright settlement; 16 cases against OpenAI await ruling
An audit finds 59.4% of the hardest problems have unsolvable or false-passing tests, inflating coding scores 5-15 points
Incidents rise to 362 logged in 2025; models show situational awareness in tests and loophole-seeking that inflates scores
Nvidia, Cisco, QIA and MGX back the Colossus compute build; xAI then leases Colossus 1 output to Anthropic
Indiana's $11bn AWS megacampus runs ~500,000 Trainium2 chips for Anthropic; the partners now target up to 5GW, with Amazon guiding to ~$200bn capex in 2026