✦Daily Brief

Today in AI

The top AI stories, summarised every morning. Free for everyone.

TodayMonday, June 1

Today in AI

AI agent bottleneck is permissions, not model performance, limiting enterprise deployment.↗
Apple working to compress Google's massive Gemini model to power iPhone Siri.↗
Startup offers free home cleaning in exchange for recording footage for robot training.↗

Sources: The Verge · VentureBeat · MIT Tech Review · TechCrunch · Ars Technica

Previous Briefs

Sunday, May 31

Today in AI

AI scammers use deepfakes of Black people to peddle Shein dropshipping on TikTok.
AI agent bottleneck is permissions, not model performance, limiting enterprise deployment.
SoftBank pledges €75 billion to build French AI data centers.
Saturday, May 30

Today in AI

AI companies pay for video footage of people doing chores to train robots.
AI agent bottleneck is permissions, not model performance, limiting real-world deployment.
Apple shrinking Google's Gemini model to run advanced AI on iPhones locally.
Friday, May 29

Today in AI

Researchers cut LLM token usage by 69.5% through automated reasoning strategy design.
Claude's new model admits mistakes more honestly than previous versions.
Apple distilling Google's Gemini model to power on-device iPhone Siri.
Thursday, May 28

Today in AI

Robinhood enables AI agents to trade stocks autonomously, risking real financial gains or losses.
MiniMax's M3 model achieves 15.6X faster long-context responses via sparse attention mechanism.
Google's AI struggles to spell basic words, including 'Google' itself.
Tuesday, May 26

Today in AI

Trump cancels AI safety EO signing after top AI firm CEOs declined attendance.
Hackers exploit chatbot personalities as AI security vulnerability grows.
Google reimagining search with agentic AI capabilities launching in 2026.
Monday, May 25

Today in AI

Hackers exploiting chatbot personalities to bypass safety guardrails and extract sensitive information.
Google's omni-modal AI model handles any input type, raising deepfake and misuse concerns.
AI agents silently causing infrastructure failures enterprises aren't tracking or detecting yet.
Sunday, May 24

Today in AI

Google's new omni-modal AI model processes any input type, showcasing breakthrough multimodal capabilities.
Top AI CEO no-show forces Trump to cancel AI safety executive order signing event.
Attackers stole npm developer accounts to bypass Sigstore's certificate security protections.
Friday, May 22

Today in AI

Elon Musk and Sam Altman's OpenAI battle escalates with latest lawsuit developments.
Alibaba's Qwen3.7-Max runs 35 hours autonomously, rivaling Claude's capabilities.
Google plans 2026 search overhaul using agentic AI to fundamentally reshape results.
Thursday, May 21

Today in AI

Elon Musk flagged as risk factor in SpaceX's upcoming IPO filing.
Google's agentic AI will remake search in 2026 with autonomous agents.
Cohere launches open Apache 2.0 Command A+ model with native citations.
Tuesday, May 19

Today in AI

Musk lost OpenAI lawsuit; jury found he waited too long to sue.
AI data centers will sustain fossil fuel demand despite solar's 2035 energy dominance.
Four AI supply-chain attacks in 50 days exposed gaps in release pipeline security.
Monday, May 18

Today in AI

Apple's revamped Siri will feature auto-deleting chats for enhanced privacy.
AI risks replacing expert knowledge it depends on to improve, warns enterprise risk analysis.
Musk and Altman's credibility clash reaches jury phase in OpenAI lawsuit.
Sunday, May 17

Today in AI

YouTube expands AI deepfake detection tool to all adult users globally.
AI is replacing expert knowledge workers it needs to learn from—major enterprise risk.
Fin launches AI agent designed solely to manage and oversee another AI agent.
Saturday, May 16

Today in AI

YouTube expands AI deepfake detection tool to all adult users globally.
ArXiv bans researchers uploading papers filled with low-quality AI-generated content.
RecursiveMAS boosts multi-agent inference speed 2.4x while cutting token usage 75%.