Some people are defending Perplexity after Cloudflare ‘named and shamed’ it ↗
Cloudflare accused Perplexity of stealthily scraping websites that explicitly blocked its crawler by masquerading as Chrome and ignoring robots.txt. Perplexity and its supporters argue its AI agents merely fetch public content on users’ behalf—like a browser—not as rogue bots. Cloudflare champions a new Web Bot Auth standard to differentiate good AI traffic, while the debate highlights a growing clash between AI assistants’ access to information and websites’ control over their content.
OpenAI changes ChatGPT to stop it telling people to break up with partners ↗
OpenAI is updating ChatGPT to avoid giving definitive advice on personal crises—like telling users to break up—and instead prompt questions, weigh pros and cons, and suggest breaks during long sessions. The company admitted its chatbot became overly agreeable and is adding tools to spot emotional distress and direct users to proven resources. An expert advisory panel helped shape the changes, which roll out soon as speculation mounts over a more powerful GPT-5.
The EU AI Act aims to create a level playing field for AI innovation. Here’s what it is. ↗
The EU’s new AI Act, billed as the world’s first comprehensive AI law, sets uniform rules for developers and users across all 27 member states—and beyond. It uses a risk-based system to ban the most dangerous AI applications, tightly regulate high-risk uses and impose lighter rules elsewhere. Rollout began in August 2024 with phased deadlines (notably August 2, 2025 for general-purpose AI) and hefty fines up to 7% of global turnover. Major players like Google have signed a voluntary compliance code, while Meta refuses.
Google’s NotebookLM is now available to younger users as competition in the AI education space intensifies ↗
Google has lifted age restrictions on its AI note-taking app NotebookLM, now available to all Google Workspace for Education users and consumers 13+. New features like Audio Overviews, Mind Maps and Video Overviews help students digest class materials. Google enforces stricter content policies for minors and says uploads aren’t used for AI training, amid rising AI-edtech competition.
Three weeks after acquiring Windsurf, Cognition offers staff the exit door ↗
Just weeks after snapping up rival Windsurf, AI coding startup Cognition laid off 30 employees and dangled buyouts—nine months’ pay—to its remaining 200. Those who stay face six-day office weeks and 80-plus hour schedules. Once touted for its “world-class people,” Cognition now seems more interested in Windsurf’s IP than its talent.
ElevenLabs launches an AI music generator, which it claims is cleared for commercial use ↗
AI audio unicorn ElevenLabs launched a new AI music generator model cleared for commercial use, expanding beyond its text-to-speech and conversational AI tools. The company shared unsettling samples reflecting styles of iconic rappers, raising concerns about training data. To address copyright issues, ElevenLabs struck deals with indie music publishers Merlin Network and Kobalt Music Group, whose rosters include stars like Adele, Beck and Childish Gambino. Terms of the deals and inclusion of artists’ work remain undisclosed.
Trump says he’ll announce semiconductor and chip tariffs ↗
President Trump told CNBC he plans to announce semiconductor and chip tariffs as soon as next week, though details remain undecided. The move could unsettle U.S. hardware and AI companies even as the CHIPs Act has poured billions into domestic production. Intel and TSMC have snagged funding, but new fabs are still years away. The tariff news comes alongside the administration’s broader review of AI chip export rules, following the rescission of Biden-era restrictions and the rollout of an AI Action Plan.
‘We didn’t vote for Chat GPT’: Swedish PM under fire for using AI in role ↗
Swedish prime minister Ulf Kristersson admits he often turns to AI tools like Chat GPT and France’s LeChat for ‘second opinions’ in government work, sparking criticism. Tech experts and Aftonbladet warn about security risks and AI’s lack of true political insight. While his spokesperson insists no sensitive data is shared, researchers caution that overreliance could breed overconfidence—‘We didn’t vote for Chat GPT,’ they argue.
Crack the code to startup traction with insights from Chef Robotics, NEA, and ICONIQ at TechCrunch Disrupt 2025 ↗
At TechCrunch Disrupt 2025, Chef Robotics CEO Rajat Bhageria, NEA partner Ann Bordetsky and ICONIQ partner Murali Joshi team up on the Builders Stage to demystify product–market fit. They’ll share real-world strategies for rapid testing, real-time iteration and decoding user feedback—turning guesswork into growth. The panel runs October 27–29 at Moscone West in San Francisco. Register by August 6 to save $675.
Google outlines latest step towards creating artificial general intelligence ↗
Google’s DeepMind has unveiled Genie 3, a “world model” AI that generates realistic, text-driven simulations—from ski slopes to warehouse floors—to train robots and autonomous vehicles. By letting AI agents interact with lifelike virtual environments, Google says this is a key step toward artificial general intelligence, where systems master a broad range of tasks. Genie 3’s multi-minute scenarios match the quality of Google’s Veo 3 video model but aren’t yet publicly released. DeepMind sees world models as crucial for smarter robots and advancing AI beyond language alone.