DeepSeek leads open-model AI for enterprises, pushing accessible, accountable AI tools amid a fast-moving AI race.
News organisations have filed a sanctions motion accusing OpenAI of concealing searchable training datasets and millions of ChatGPT logs that could show whether the company used copyrighted journalism. Depositions have revealed OpenAI held large, de‑identified log samples and internal tools to detect regurgitation; plaintiffs say the sample OpenAI produced was unusable.
Moonshot AI has released Kimi K3, a 2.8 trillion-parameter open-weight model that outperforms many rivals on coding benchmarks and is set to be released as open weights on July 27. The launch challenges U.S. dominance in frontier AI, lowers costs, and intensifies regulatory and competitive pressures.
DeepMind CEO has proposed a FINRA-like AI standards body to rigorously test frontier models before release, aiming to curb cybersecurity, bio and nuclear risks. The push is advancing in Washington amid broader U.S.-led regulatory debates and international interest.
Open-source AI models are expanding their footprint as enterprises shift away from exclusive frontiers. Chinese models are rising in popularity on data platforms while European and open-weight strategies gain traction, signaling a diversified AI ecosystem. The developments come as startups secure funding and hardware suppliers scale to meet production needs.
AI industry leaders are facing intensified threats as public sentiment turns against AI. Attacks on executives and facilities have prompted heightened security measures, while policy discussions on wealth sharing and regulation intensify.
The White House has issued a memo saying foreign actors, principally based in China, have been running industrial-scale campaigns to "distil" US frontier AI systems by using proxy accounts and jailbreaking techniques to extract capabilities. The administration has said it will share intelligence with US AI firms and explore measures to punish offenders ahead of a planned US–China summit.
Nvidia CEO Jensen Huang has joined President Trumps delegation to China and has been pictured in Beijing; the trip has been focusing on trade, AI export controls and Iran. China has not approved any purchases of Nvidias H200 chips and is continuing to push domestic chip development while U.S. export controls remain in place.
OpenRouter and Concentrate AI are expanding AI-model routing, driving demand for cheaper models and cost controls as Anthropic and OpenAI race to go public. The landscape shifts as buyers seek value and governments debate how to share benefits from AI progress.
Stocks rebound after major tech IPOs, with AI-focused names leading gains while volatility remains elevated. SpaceX’s blockbuster listing prompts index-weight discussions; traders weigh how dispersion and inflation signals might steer the market ahead of more mega-IPOs.
A coalition of U.S. state attorneys general has subpoenaed OpenAI for internal documents on advertising, user engagement, handling of health and consumer data, and protections for minors and seniors. OpenAI has said it will "engage constructively," highlighted new safeguards in ChatGPT and is cooperating with investigators while facing related lawsuits and regulatory pressure.
The White House has restricted access to Anthropic’s Mythos 5 and Fable 5 for foreign nationals, citing national security. Anthropic has suspended access to these models, creating winners for open-weight providers like Mistral and DeepSeek and prompting renewed calls for AI sovereignty in Europe. The debate centers on control versus access as nations weigh strategic dependencies on US AI infrastructure.
OpenRouter data shows open-source AI models like GLM-5.2 are gaining traction against top US models, offering cost advantages and enterprise use. OpenRouter traffic is rising while concerns over safety, governance, and regulatory exposure accompany the shift as firms weigh token costs and performance.
GPTZero, the AI-detection startup founded by Princeton graduate Edward Tian, has been acquired by Superhuman. The deal values GPTZero at over $88 million, with 19 million+ users and $30 million in annual recurring revenue; the integration will bolster authenticity tools across Superhuman’s platform while keeping GPTZero as a stand-alone product.
Anthropic has alerted lawmakers to a campaign by operators linked to Alibaba’s Qwen lab that allegedly carried out 28.8 million exchanges with Claude across nearly 25,000 fraudulent accounts between April 22 and June 5, aiming to extract its capabilities. The company says the activity is the largest known distillation attack to date and calls for penalties and stronger safeguards.
South Korea has accelerated its semiconductor push, pledging hundreds of trillions of won in memory fabs and AI data centers. President Lee Jae-myung frames the plan as national survival, aiming to double memory capacity within five years. Samsung and SK Hynix pledge multi‑trillion investments, with broader plans to build new fabs and hubs in the southwest.
Meta is pursuing a cloud compute business, potentially selling access to excess AI compute power and/or hosted AI models. The move follows high AI infrastructure spending and could reshape the company’s revenue flow beyond advertising. Investors have reacted positively, while rivals and early pilots offer a benchmark for profitability.
Menlo Ventures has backed Anthropic since 2023, turning a $4.1 billion pre-money valuation into a potential multi-billion-dollar windfall as Anthropic prepares to go public. The investment has reshaped Menlo, contributing to a new $3 billion fund and elevating Anthropic as a major challenger to OpenAI in AI safety and capability.
Enterprises are shifting from chasing the top model to integrating best-fit open-weight options, aiming to cut costs while preserving performance. Open-weight models are gaining traction as a flexible backbone for task-specific systems, with large labs facing pressure as organizations route work to cheaper, capable engines.
The AI industry has reached a tipping point as hyperscalers project rising returns from chips and data centers. New analyses forecast trillions in 2026 revenue to justify the AI build-out, even as token costs and cheaper models pressure profitability.
China’s Cyberspace Administration has approved Apple’s AI services for use in the country, following a deal to integrate Alibaba’s Qwen AI into Apple’s operating systems. Apple’s China sales have risen, and Baidu is reportedly collaborating on Apple Intelligence features for Chinese users. The approval sets the stage for broader AI integration in China.
Xi Jinping has urged international cooperation on artificial intelligence, arguing it should not be dominated by one country. At the World AI Conference in Shanghai, Beijing has highlighted WAICO and pledged to expand AI capacity-building for developing nations, while stressing that AI must remain under human control and guided by a people-centered approach.