Anthropic and DoD clash over Claude use in surveillance and weapons
TechCrunch · February 15, 2026
Anthropic is reportedly in a dispute with the Pentagon over whether Claude can be used for mass domestic surveillance and autonomous weapons. The standoff highlights how AI vendors’ safety policies are colliding with national security demands, setting a precedent for what “responsible use” means in government contracts and how enforceable those limits really are.
Read full article →Hollywood pushes back on Seedance 2.0 over copyright concerns
TechCrunch · February 15, 2026
Hollywood groups are criticizing the new Seedance 2.0 AI video generator, arguing it’s rapidly become a tool for blatant copyright infringement. The backlash signals escalating tension between generative video platforms and rights holders, and could accelerate legal and policy pressure on how these models are trained and used.
Read full article →NPR’s David Greene sues Google over NotebookLM’s podcast voice
TechCrunch · February 15, 2026
Former NPR “Morning Edition” host David Greene is suing Google, alleging the male podcast voice used in NotebookLM was modeled on his own voice. The case spotlights growing legal pressure around AI voice cloning and the need for clearer consent, licensing, and provenance rules for synthetic media in consumer products.
Read full article →Blackstone eyes $1.2B for Neysa to scale India’s AI compute
TechCrunch · February 16, 2026
Blackstone is backing Indian AI infrastructure startup Neysa with financing of up to $1.2B as India accelerates efforts to build domestic compute capacity. Neysa plans to deploy 20,000+ GPUs over time, aiming to meet rising local demand for training and inference without relying on overseas cloud supply. The deal highlights how capital and policy are converging to reduce GPU scarcity and strengthen India’s AI stack at home.
Read full article →Ricursive Intelligence hits $4B valuation after $335M raise in 4 months
TechCrunch · February 16, 2026
TechCrunch reports Ricursive Intelligence raised $335M at a $4B valuation just four months after launch, with investor demand driven largely by the founders’ reputations in the AI community. The deal underscores how elite AI talent can compress fundraising timelines and valuations, even before a startup has a long operating track record.
Read full article →Fractal Analytics’ IPO debut highlights India’s AI valuation jitters
TechCrunch · February 16, 2026
Fractal Analytics, billed as India’s first AI company to go public, had a subdued market debut as broader selling pressure in Indian software stocks dampened enthusiasm. The reception suggests investors are still cautious about pricing AI-heavy business models, making it a key read on how India’s public markets may value and fund AI firms going forward.
Read full article →Mistral AI acquires Koyeb to accelerate its cloud platform push
TechCrunch · February 17, 2026
Mistral AI is buying Paris-based Koyeb, a startup focused on simplifying deployment and scaling of AI applications while managing the underlying infrastructure. It’s Mistral’s first acquisition and signals a move beyond model development toward owning more of the cloud stack—potentially improving performance, cost control, and enterprise-ready deployment options.
Read full article →Adani commits $100B to 5GW AI data center buildout in India
TechCrunch · February 17, 2026
Adani says it will invest $100B to develop up to 5 gigawatts of AI-focused data center capacity in India, with plans tied to partnerships including Google, Microsoft, and Flipkart. The scale signals a major push to expand India’s compute and cloud infrastructure as global AI demand strains power and supply chains, potentially reshaping where large-model training and inference can be hosted.
Read full article →Report: Apple is developing three AI-focused wearables
TechCrunch · February 17, 2026
Apple is reportedly working on a trio of AI wearables as competition in dedicated AI hardware accelerates. If the plans pan out, it signals Apple is exploring new form factors beyond the iPhone to run AI features closer to the user—raising the stakes for on-device intelligence, privacy, and the broader wearables ecosystem.
Read full article →World Labs raises $1B, teams with Autodesk to embed world models in 3D
TechCrunch · February 18, 2026
World Labs has secured a $1B round, including $200M from Autodesk, alongside a partnership to integrate its “world models” with Autodesk’s 3D tools. The effort starts with entertainment workflows, aiming to speed up scene creation and iteration by pairing generative world building with established design and production pipelines.
Read full article →Microsoft: Office bug let Copilot summarize confidential customer emails
TechCrunch · February 18, 2026
Microsoft disclosed an Office bug that caused Copilot to read and summarize some paying customers’ confidential emails, bypassing intended data-protection policies. The incident highlights how small integration flaws in AI-enabled productivity suites can turn into material data exposure risks. For enterprises, it’s a reminder to scrutinize AI access boundaries, auditing, and rollout controls—not just model behavior.
Read full article →Sarvam ships 30B/105B open-source models plus speech and vision stack
TechCrunch · February 18, 2026
Indian AI lab Sarvam released a new open-source lineup spanning 30B and 105B parameter LLMs, plus dedicated text-to-speech, speech-to-text, and document-vision models. The move signals a push to make high-end, full-stack AI capabilities available outside closed platforms, giving developers and enterprises more control over deployment, cost, and data governance.
Read full article →Report: OpenAI nears $100B raise at $850B+ valuation
TechCrunch · February 19, 2026
TechCrunch reports OpenAI is close to finalizing a $100B funding deal that would value the company at more than $850B, with backers said to include Amazon, Nvidia, SoftBank, and Microsoft. If confirmed, it would be one of the largest private financings ever and signals escalating competition to secure AI compute, talent, and distribution. The investor mix also underscores how strategically important OpenAI’s models have become across cloud, chips, and enterprise software.
Read full article →Reliance bets $110B on AI, building multi-GW data centers in Jamnagar
TechCrunch · February 19, 2026
Reliance says it will invest $110B in AI and is already building multi-gigawatt AI data centers in Jamnagar, with more than 120 MW slated to come online in 2026. The move signals India’s push to grow domestic AI compute capacity at hyperscale, potentially reshaping regional cloud/AI infrastructure and competitive dynamics for global providers.
Read full article →OpenAI partners with Tata for 100MW India AI capacity, targets 1GW
TechCrunch · February 19, 2026
OpenAI is reportedly securing 100MW of AI data center capacity in India through Tata, with plans that could scale to 1GW. It also intends to open new offices in Mumbai and Bengaluru later this year, signaling a deeper push into India’s fast-growing AI market. The move matters as it localizes compute and operations, improving latency, compliance options, and supply for Indian customers amid global GPU constraints.
Read full article →G42 and Cerebras to bring 8 exaflops of AI compute to India
TechCrunch · February 20, 2026
UAE-based G42 is partnering with U.S. chipmaker Cerebras to deploy a new AI compute system in India totaling 8 exaflops. The move expands India’s access to cutting-edge training and inference capacity and signals continued competition to place large-scale AI infrastructure outside the U.S. and China.
Read full article →Google’s Gemini 3.1 Pro tops benchmarks, targets harder workloads
TechCrunch · February 20, 2026
Google has unveiled Gemini 3.1 Pro, claiming new record benchmark scores and positioning it as a more capable model for complex, multi-step work. If the gains translate beyond benchmarks, it could strengthen Google’s hand in enterprise AI—especially for developers building agentic workflows and tool-using applications where reliability and reasoning depth matter.
Read full article →ggml.ai joins Hugging Face to boost long-term Local AI development
Hacker News · February 20, 2026
The ggml.ai team behind the ggml/llama.cpp ecosystem is joining Hugging Face, aiming to secure long-term maintenance and progress for local, on-device LLM tooling. For developers and users running models outside the cloud, the move could mean more stable funding, better infrastructure, and closer integration with Hugging Face’s open-source stack—while keeping momentum in the fast-moving “local AI” space.
Read full article →Google VP: LLM wrappers and AI aggregators face a squeeze
TechCrunch · February 21, 2026
A Google VP cautions that two common gen‑AI startup models—LLM “wrappers” and AI aggregators—may struggle to survive as core model providers add features and compete directly. With limited defensibility and shrinking margins, these businesses risk being commoditized unless they build differentiated data, workflows, or distribution. The warning underscores how quickly the stack is consolidating and why moats matter more than ever in AI.
Read full article →OpenAI Weighed Alerting Police Over Suspected Shooter’s ChatGPT Use
TechCrunch · February 21, 2026
TechCrunch reports that OpenAI internally debated contacting law enforcement after monitoring tools flagged a user’s chats describing gun violence. The episode highlights how AI companies are operationalizing safety monitoring—and the hard trade-offs between preventing harm, user privacy, and when a “credible threat” threshold should trigger real-world escalation.
Read full article →zclaw squeezes a personal AI assistant into 888 KB on an ESP32
Hacker News · February 21, 2026
An open-source project called zclaw demonstrates a tiny “personal AI assistant” that fits in under 888 KB and runs directly on an ESP32 microcontroller. It’s a practical example of pushing AI-like functionality to ultra-low-cost edge hardware, highlighting how embedded assistants can work with tight memory and compute budgets without relying on always-on cloud services.
Read full article →