Homechevron_rightArchives

Archives

Explore Stories by Date

25 Jun 2026

Chinese AI models close gap with Anthropic and OpenAI as GLM-5.2 tops open-weight rankings

Beijing-based Z.ai released GLM-5.2, an open-weight AI model that ranks fourth globally on intelligence benchmarks while costing a fraction of Western alternatives. Trained entirely on Huawei Ascend chips, the model arrives as US export controls on AI face scrutiny and Anthropic's Fable 5 remains restricted, shifting the competitive landscape.

10 Sources

Sega's Sonic the Hedgehog ARG sparks fury as fans discover hidden AI data training clause

Sega launched a Sonic the Hedgehog alternate-reality game for the franchise's 35th anniversary, but fan excitement turned to outrage when participants discovered the terms of service require consent for AI data training. The Chaos Hunt ARG collects names, birthdates, zip codes, and phone numbers through marketing platform Community, which explicitly states user data may train proprietary AI models and be shared with third parties.

3 Sources

Alan raises €480M led by Prosus at €5.5 billion valuation for AI-driven prevention insurance

French insurtech Alan has secured €480M in Series G funding led by Prosus at a €5.5 billion valuation, just three months after its previous round. The Paris-based company is reshaping health insurance with AI-driven prevention, bundling health cover, care navigation, and wellbeing services into a single platform that acts before illness strikes.

2 Sources

AMD brings FSR 4.1 to RDNA 3 graphics cards ahead of schedule, updates SDK with ray tracing boost

AMD has released FSR Upscaling 4.1 for Radeon RX 7000 series cards earlier than expected, originally scheduled for July. The AI-powered upscaling tech was previously exclusive to RX 9000 cards. Meanwhile, the FSR SDK 2.3.0 update delivers FSR Ray Regeneration 1.2.0 with quality improvements and expanded denoising capabilities for game developers.

2 Sources

Anthropic's hiring spree targets Australia and Japan for AI data center expansion

Anthropic is ramping up its AI infrastructure in the Asia-Pacific region with 13 new compute department roles, eight focused on Australia and Japan. The expansion comes as soaring demand for its Claude models strains infrastructure, with the company's revenue run-rate hitting $47 billion in May. But challenges including copyright laws and power access could complicate the buildout.

2 Sources

Anthropic's White House Negotiations Improve After CEO Dario Amodei Replaced by Co-Founder

Anthropic CEO Dario Amodei has been sidelined from critical White House negotiations over export controls on the company's advanced AI models. Co-founder Tom Brown and public policy chief Sarah Heck are now leading talks with the Trump Administration, with sources indicating discussions are progressing better since the leadership change.

2 Sources

Runpod raises $100M at $1 billion valuation as AI developers scramble for computing power

Runpod has raised $100M in Series A funding led by Summit Partners, achieving a $1 billion valuation—a tenfold jump in under two years. The cloud startup rents AI computing power to over one million developers, offering a full-cycle platform for experimentation, training, and deployment. The company rejected buyout offers worth more than $500M to stay independent.

2 Sources

Anthropic hires Orange's AI chief Steve Jarrett as European expansion accelerates

Steve Jarrett, who served as Orange's Chief AI Officer since 2019, is joining Anthropic in Paris to help adapt its Claude AI models for European and African markets. The move comes as the AI startup opens its sixth European office in Milan and plans to triple its international workforce ahead of a planned IPO this year.

3 Sources

EU joins Pax Silica despite France's objections, raising questions about tech sovereignty

The European Union has joined Pax Silica, a US-led initiative to secure AI chip supply chains and coordinate export controls against China. The decision comes just weeks after Brussels unveiled a tech-sovereignty agenda aimed at reducing dependence on foreign suppliers, creating tension between autonomy and practical cooperation on advanced semiconductors.

2 Sources

IBM unveils sub-1 nanometer chip technology with 100 billion transistors for AI data centers

IBM revealed the world's first sub-1 nanometer chip technology, packing nearly 100 billion transistors onto a fingernail-sized die using a new nanostack architecture. The breakthrough delivers up to 50% higher performance and 70% greater energy efficiency compared to IBM's 2-nanometer chips, with production expected within five years through partners like Rapidus.

12 Sources

Tim Sweeney calls Valve's Steam AI disclosures irresponsible as Unreal Engine doubles down on AI

Epic Games CEO Tim Sweeney has sharply criticized Valve's requirement for AI disclosures on Steam, calling it irresponsible and claiming it makes success much harder for game developers. The comments come as Epic unveils AI integrations in Unreal Engine 5.8 and the upcoming Unreal Engine 6, intensifying the debate over AI in game development and transparency in the gaming industry.

4 Sources

Indian Government Set to Acquire Minority Stake in Sarvam AI Through IndiaAI Mission Support

The Indian government could secure a 1-2% stake in AI startup Sarvam as part of its $300 million funding round, marking a rare instance of direct equity holding in a private AI company. The stake stems from compute infrastructure provided under the IndiaAI Mission, where compulsorily convertible debentures will convert to equity at a $1.5 billion valuation.

2 Sources

Qualcomm unveils High-Bandwidth Compute architecture to break AI memory wall with 6x efficiency

Qualcomm introduced its High-Bandwidth Compute (HBC) architecture at Investor Day 2026, positioning itself as a serious contender in the AI data center market. The technology stacks compute beneath DRAM to tackle the memory wall bottleneck, claiming 6x higher bandwidth-per-watt compared to HBM and 200x capacity versus on-chip SRAM. The first HBC-powered AI250 accelerators are expected by mid-2027.

4 Sources

Xiaomi's HarnessX rewrites AI scaffolding mid-task, proving agent harness trumps model size

Xiaomi researchers introduced HarnessX, a framework that autonomously rewrites AI scaffolding during task execution, delivering an average +14.5% performance gain across 15 model-benchmark combinations. For smaller models like Qwen3.5-9B, gains reached +44% on embodied planning tasks, demonstrating that scaling foundation models isn't the only path to more capable AI agents.

2 Sources

Hollywood workers quietly train AI models as entertainment jobs dry up and survival takes priority

Three years after the 2023 strikes raised alarms about AI replacing entertainment workers, some of those same workers are now training the technology that worries them. As film and TV jobs grow harder to find, writers, editors, and executives across Hollywood are quietly taking gig work in Reinforcement Learning from Human Feedback to pay the bills.

2 Sources

Get all updates into your inbox 💌, Stay Up-to-date for Free 🥳

Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Follow topics that matter to you and stay ahead.

Explore

News Categories

Technology Business Policy Startups Health Science Entertainment

Terms Privacy Content Contact Us