AI
AIMIT’s city cameras can count emissions and raise a surveillance problem
A new book argues computer vision could make streets measurable at scale, while warning that more cameras carry costs.
AIGoogle proposes cloud memory that even Google says it cannot read
A new Private AI Compute design would store an assistant's long-term context in the cloud while keeping decryption keys on a user's devices.
AIClaude agents traded 201 people's books. Reading their tastes was the hard part
Anthropic's book-swap experiment suggests an AI negotiator may bargain capably yet still misunderstand the person it represents.
AIMicrosoft puts an agentic security operations center in Defender preview
Microsoft’s new ISOC preview combines security data and controls inside Defender for humans and AI agents. The company has not published independent evidence that it improves outcomes.
AIESA and Mistral set a framework for AI in European space work
The European Space Agency and Mistral signed a letter of intent to explore AI for Earth observation and engineering. It names possible safeguards, but commits neither side to a deployed system.
AIMicrosoft’s science agent helps probe wastewater viruses, with caveats
A Microsoft Discovery case study describes AI-assisted analysis of wastewater sequencing data. The work remains exploratory, with candidate signals requiring human review and laboratory validation.
AIMicrosoft’s RetroChimera plans molecule recipes, but chemists still need to test them
The AI retrosynthesis system won expert approval in nine of ten benchmark routes, but its own team warns that predictions can hallucinate.
AIIBM study finds workers worry AI erodes judgment as review work grows
In a global survey, 60% of employees worried AI was eroding their skills, while HR chiefs described new invisible work from checking AI output.
AIClaude Opus 5.5 cuts typical agent costs by 40%, Anthropic says
Anthropic pairs a cheaper flagship model with stronger agent safeguards, but its real-world cost and safety claims still need independent testing.
AIGoogle.org puts $4 million behind free AI training for educators
Digital Promise will expand teacher and community-college training, then publish guidance on what helps faculty use AI well.
AIGrok 4.7 rolls into GitHub Copilot with usage-based billing
GitHub is adding xAI’s model across editors, CLI and cloud agent, while administrators retain model policy controls.
AIAI usage reaches 18.8% worldwide as the regional gap widens
Microsoft’s new diffusion estimate shows broad AI adoption gains but a growing divide between the Global North and South.