AI News from TechCrunch AI

Latest coverage from TechCrunch AI, summarized and scored for signal.

• AI isn’t close to curing cancer. This startup says it knows what it will take. — Biotech startup Vivodyne has opened what it calls the world's largest "human data center" near San Francisco, using HIVE robotic labs to generate human tissue data that AI drug-discovery models currently lack — pushing back against industry hype that AI is already close to curing cancer.
• Relativity Networks raises $22 million to bring a faster kind of fiber to data centers — Relativity Networks raised $22M and secured a $40M follow-on order from an unnamed hyperscaler to deploy hollow-core fiber in data centers—a technology transmitting data 30% faster than conventional glass-core fiber by routing light through a vacuum channel.
• Cursor capitalizes on Github frustration, launches rival hosting platform — Cursor, now part of SpaceXAI, launched Origin — a GitHub-competing code hosting platform — on the same day GitHub suffered a 6-hour worldwide outage with a nearly 20% error rate, intensifying pressure on a service with 180 million developers and 257 outages in the past year.
• OpenAI institutes new safeguards after Hugging Face breach — OpenAI announced new security practices — including real-time behavioral monitoring at 20% compute overhead, tighter network isolation, and 30-minute alert targets — following July's Hugging Face breach. Its largest planned frontier RL run remains on hold while smaller-scale evaluations validate the safeguards.
• Etched’s valuation doubles to $21B in a month — Etched raised $700M at a $21B valuation — led by Jane Street after the quant fund deployed and tested an Etched inference cluster in its own data center. The valuation tripled from $5B in December and nearly doubled from $10.3B in July, driven by validated hardware that combines a custom low-voltage prefill chip with cluster-scale shared memory for decode.
• Why Apple’s camera-equipped AirPods may not be the ‘pervert pods’ consumers fear — Apple's leaked AirPods Pro 4 will include cameras, but code in macOS 26.7 RC shows they'll function exclusively as low-resolution eyes for Siri's Visual Intelligence — unable to record photos or video — a design choice that avoids the privacy backlash plaguing Meta's Ray-Bans.
• Warp’s new system is an out-of-the-box software factory for AI development — Warp launched Warp Factories, an out-of-the-box infrastructure layer that automates the software factory model — agent loops spanning triage, spec, implementation, review, and verification — targeting smaller companies that can't build this architecture from scratch.
• OpenAI launches a safer ChatGPT for teens — years after teens started using it — OpenAI launched ChatGPT for Teens with age-appropriate safety guardrails, a Study Mode that guides learning rather than providing answers, parental controls with Quiet Hours, and a CodeAI partnership — responding to lawsuits and mental health concerns over teens using unguarded AI chatbots.
• Perplexity’s free AI offer left it with millions more users in India — Perplexity's 12-month free Pro giveaway to Airtel's 360 million Indian subscribers drove 56 million app downloads over 7 months — and India revenue rose ~60% after the offer ended, providing rare concrete data on whether freemium AI bundling converts to paying users in emerging markets.
• Anthropic’s annualized revenue surges to $65B — Anthropic's annualized revenue run rate hit $65B at end of July 2026, up from $47B in May and $9B at end of 2025 — investors expect $100-120B by year-end as the company files confidential IPO paperwork targeting a $2T public valuation.
• AI automation startup Relay shuts down, staff joins Google’s Chrome team — Relay, an AI workflow automation startup aiming to be the new Zapier, is shutting down September 14 as founder Jacob Bank rejoins Google as VP of Product for Chrome, teasing ambitious AI-in-browser integrations.
• Amazon, once an online bookseller, is destroying rare books to train AI models — Amazon is physically destroying rare books — cutting off spines and scanning them — to generate training data for its LLMs, according to 404 Media, which tracked a book to Amazon's VGT3 facility in Las Vegas.
• Groq raises $350M to fuel its pivot from AI chips to neocloud — Groq raises $350M at a $3.5B valuation — down 49% from its September 2025 peak — as the former LPU chipmaker completes its pivot to an Nvidia-powered neocloud with 13 data centers serving 6 million developers and enterprises.
• Nvidia investing $1.5B in SoftBank data center developer behind OpenAI project — Nvidia is investing $1.5B in SB Energy—the SoftBank-backed developer building OpenAI's Ports-Pike data center near Cincinnati—plus providing up to $105B in credit, locking itself in as the sole compute supplier for a facility that may scale to 8 gigawatts.
• Wispr raises $280M at $2B valuation as it looks beyond dictation — Wispr raised $280M Series B at a $2B valuation led by Menlo Ventures and is launching Canto, a new speech model targeting a drop in dictation error rates from 30% to under 10%, while expanding into meeting notes to compete with Granola and Fireflies.
• Stripe will reportedly acquire AI gateway startup OpenRouter for $7B+ — Stripe has agreed to acquire OpenRouter — an AI model gateway serving 8 million users with access to 400+ models — for more than $7 billion, marking one of the largest AI infrastructure acquisitions on record.
• Why people aren’t buying Mark Zuckerberg’s AI future — TechCrunch's Equity podcast dissects Zuckerberg's 6,500-word AI manifesto, questioning whether Meta can deliver on personal AI promises given its social media track record — its essay describes two new models: Glimmer (always-on personal AI) and Muse Spark (premium cloud tier).
• Anthropic CEO says AI backlash is ‘fundamentally a crisis of trust’ — Dario Amodei rejects investor Gavin Baker's claim that AI safety warnings fueled the industry backlash, framing it instead as decades-long distrust of institutions — and says the real failure is AI companies not yet delivering on their promised benefits to society.
• Woman claims her stepfather used Grok to transform childhood photo into explicit imagery — A fourth victim joined the Tennessee CSAM class-action against xAI, alleging Grok was used to generate 7,000+ explicit images from a childhood photo — intensifying legal pressure on AI companies that lack guardrails for image generation.
• Anthropic shares more details about how Claude’s new watermarks will work — Anthropic adopted Google DeepMind's SynthID-Text watermarking for Claude to comply with the EU AI Act, embedding invisible statistical patterns in word choices—with a public detection API planned, though a full word-for-word rewrite can defeat it.
• SpaceX officially closes its Cursor acquisition — SpaceX officially closed its $60 billion acquisition of AI coding tool Cursor, giving the startup access to SpaceX's GPU fleet — currently rented by Anthropic and Google — and completing Elon Musk's vertical integration of AI infrastructure alongside xAI, acquired earlier this year.
• How to tell if your AI platforms’ accounts have been hacked — TechCrunch details how to detect and revoke unauthorized sessions on ChatGPT, Claude, and Perplexity — including how Claude's email-link-only login (no MFA, no password reset) limits account recovery options compared to competitors.
• Google will now allow users to remove visible watermark from its AI generations — Google makes visible AI watermarks optional in Gemini and Flow while keeping invisible SynthID and C2PA metadata — and is open-sourcing Credentio, a new library letting developers embed local watermark verification into their own apps.
• Does Mark Zuckerberg really believe AI is ‘for everyone’? — Meta released Glimmer as an open-weight model for local download alongside closed-API Muse Spark, while Zuckerberg published a 6,500-word manifesto on AI democratization — TechCrunch's Equity podcast questions whether the dual-tier release matches the 'AI for everyone' rhetoric.
• Kog is going deeper to squeeze more inference out of GPUs — French startup Kog hit 3,000 tokens/second on standard H200/MI300X GPUs in a software-only demo, generated 200 business leads from a Hacker News front page, and is now racing to scale from a 2B-param proof-of-concept to full-size LLMs.