AI News for January 24, 2026 — 7min.ai

[NEWS] Yann LeCun's AMI Labs officially launches to build world models, reportedly in talks at $3.5 billion valuation: Yann LeCun's new Paris-based startup AMI Labs has launched with a mission to build "world models" that understand the real world — an approach he argues will succeed where LLMs fundamentally cannot.
[NEWS] AI luminaries clash at Davos over how close human-level intelligence really is: Demis Hassabis, Dario Amodei, and Yann LeCun offered starkly different views on AGI timelines at the World Economic Forum.
[NEWS] Meta blocks teens from accessing AI characters globally following problematic chat reports: Meta is shutting down access to AI characters for all teenage users worldwide across its apps, effective in the coming weeks.
[NEWS] Google DeepMind goes on acquisition spree with three AI deals in one week: Google DeepMind closed deals with three AI companies in a single week: acquiring 3D AI startup Common Sense Machines, hiring Hume AI's CEO and engineers in a licensing deal to improve Gemini's voice capabilities, and investing in Japan's Sakana AI at a $2.5 billion valuation.
[NEWS] Inside the Thinking Machines meltdown: Months of secret talks with Altman preceded co-founder exodus: New reports reveal the split at Mira Murati's Thinking Machines Lab had been brewing for months.
[RESEARCH] AI safety defenses collapse under scrutiny in new research from OpenAI, Anthropic, and Google DeepMind: A joint research paper from OpenAI, Anthropic, and Google DeepMind reveals that current AI safety measures are far more fragile than previously understood.
[NEWS] OpenAI's API business grew by $1 billion ARR last month: OpenAI CEO Sam Altman announced the company added over $1 billion in annual recurring revenue from its API business in just the past month.
[NEWS] IMF chief warns AI will be 'tsunami' hitting labor market, with young workers worst affected: IMF Managing Director Kristalina Georgieva told Davos delegates that AI will transform 60% of jobs in advanced economies and 40% globally — either enhancing, eliminating, or fundamentally changing them.
[RESEARCH] GPT-5.2 Pro sets record on FrontierMath benchmark, solving problems no AI model could before: OpenAI's GPT-5.2 Pro achieved 31% on the hardest tier of the FrontierMath benchmark — a major leap from Gemini 3 Pro's previous best of 19%.
[NEWS] TikTok finalizes US spinoff deal with Oracle, MGX, and Silver Lake as managing investors: TikTok has closed its deal to spin off parts of its US business in a joint venture with Oracle, Silver Lake, and Abu Dhabi's MGX, each owning 15%.

About 7min.ai · Subscribe to newsletter