June 12, 2026
Today's top tech stories, deduped across the newsletters I read and briefly summarized. Click a source to open the original article.
AI
Dario Amodei: AI progress is outpacing lawmakers
Anthropic CEO Dario Amodei warns that rapid technological progress is outrunning legislative capacity and calls for mandatory safety testing and government oversight of frontier models to prevent catastrophic risks in cybersecurity and biology. He proposes a regulatory model inspired by the FAA.
OpenAI weighs deep price cuts ahead of expected Anthropic competition
OpenAI is considering significant token price cuts to counter anticipated reductions from Anthropic, as enterprise customers push back on high AI costs. Both companies are already losing billions of dollars, and a price war would likely compress margins further.
EU orders Meta to open WhatsApp to rival AI chatbots
The EU has ruled that Meta is abusing its dominant position by blocking third-party AI chatbots from the WhatsApp Business API, and is requiring Meta to open access for free. Meta banned third-party AI chatbots from the platform in October last year.
Anthropic's Fable model draws fire over heavy-handed filters and system prompt leak
Security researchers are frustrated that Anthropic's public Fable model blocks legitimate cybersecurity requests like code review and blog analysis, silently falling back to Claude Opus 4.8. Separately, the model's full system prompt — approximately 120,000 characters — has been leaked online.
Microsoft launches MAI-Image-2.5, now ranked No. 2 on image-editing leaderboard
Microsoft has released MAI-Image-2.5, its strongest image model to date, which ranks second on Arena's image-editing leaderboard ahead of GPT-Image-1.5. The model supports precise, localized edits including object replacement, background cleanup, and face identity preservation.
Security
Ivanti patches critical Sentry bugs rated 10.0 and 9.9
Ivanti is urging customers to patch immediately after two critical Sentry flaws were disclosed: one lets unauthenticated attackers execute code as root via an exposed Tomcat API, while the other allows remote attackers to create rogue admin accounts on affected gateways.
Anthropic maps AI-enabled cyber threats using MITRE ATT&CK
Anthropic analyzed 13,873 technique observations from 832 banned accounts and found that medium-or-higher-risk actors rose from 33% to 56% in one year, with most using AI for early-stage capability development and defense evasion. The findings are visualized in a new LLM ATT&CK Navigator.
Dev
Lovable hits $500M ARR with just 146 employees
Vibe-coding platform Lovable reports $500 million in annual recurring revenue with only 146 employees, claiming 80% of its builders are non-technical users. The platform hosts over 50 million projects that receive 720 million monthly visits.
IT
Broadcom, Apollo, and Blackstone launch $35 billion AI infrastructure platform
The three companies are launching the AI XPV Platform with $35 billion in backing, targeting frontier AI labs and designed to deliver more than 20 gigawatts of compute capacity through 2028.
OpenAI weighs 10 GW Ohio data center campus backed by NVIDIA
OpenAI is considering a 20-year lease for a 10 GW data center campus in Ohio backed by NVIDIA, with operations expected to begin in 2028 and OpenAI controlling the computing equipment on site.
NVIDIA Quantum InfiniBand adds one-click multi-tenant fabric security
NVIDIA Quantum InfiniBand now offers intent-based security profiles in Unified Fabric Manager (UFM) that enable multi-tenant fabric security in a single click. The feature simplifies security configuration for large-scale AI infrastructure deployments.
Other
SpaceX IPO more than four times oversubscribed, set to be the largest ever
Demand for SpaceX's IPO has exceeded available shares by more than four times, with 555.6 million shares priced at $135 each. The offering is expected to rank as the biggest IPO in history.