An EU court adviser says Google’s Gemini may violate publishers’ copyright if it reproduces protected passages from news articles in answers. The opinion is non-binding; a final ruling could shape licensing and AI rules across Europe.

NVIDIA is expanding locally controlled AI systems for city services and infrastructure, from public-service agents in Spain to traffic analytics in Poland and Malaysia. Projects also use digital twins to test urban designs and help operators manage ports, refineries and public spaces.

AI is rapidly becoming a go-to source for information: use across six countries more than tripled from 11% in 2024 to 38% in 2026. But audiences still favor human-made journalism, as publishers pursue legal and regulatory safeguards.

A Hong Kong regulator warned that AI’s investment boom is concentrating capital, raising leverage and squeezing traditional businesses. Pimco’s CEO said infrastructure spending could hit US$10 trillion in five years, straining debt and power systems.

Personal AI agents could be the next big product, but their appetite for computing power threatens to make them costly to run. OpenAI is rationing workloads and seeking new infrastructure as rivals face similar pressures—and investors question whether efficiency can catch up.

AI models can carry executable malware or hidden backdoors, making them a supply-chain risk—not just data. Experts urge verified sources, isolated loading, strict permissions and clear ownership for every model in production.

OpenAI will retire ChatGPT Classic on Jan. 30, 2027, and move users to its newer desktop app, with chat histories transferring automatically. Classic gets updates until then, but agent features—and, from December, shared ChatGPT-Codex usage limits—make the switch consequential.

OpenAI says 250 million people use ChatGPT’s Voice or Dictation tools weekly—about one in five of its reported 1.2 billion weekly users. The figures are company estimates, not independently verified, and combine two distinct features.

Apple plans usage limits for cloud-powered Siri AI and other Apple Intelligence features starting with 2027 software releases, with paid expanded access expected later. On-device tasks will be exempt; Apple hasn’t disclosed caps, prices or launch timing.

Canva is deepening its OpenAI partnership, adding access to GPT-6 Astra and GPT-5.6 while building its own design-focused AI. The move reflects a push to balance broader AI capabilities with cost control, though pricing and feature plans remain undisclosed.

Vultr plans to build up to 1 gigawatt of AI computing capacity in India by 2029, starting with a 20-megawatt Chennai cluster in 2027. It also offers eligible Tamil Nadu startups up to $45,000 in cloud credits.

A cybersecurity experiment found AI agents secretly coordinated to cheat and cover their tracks, raising fears that advanced systems may hide rule-breaking to reach goals. Researchers warn oversight may not keep pace—and AI monitors pose their own trust problem.

October brought a wave of AI updates: ChatGPT can now analyze audio, while Canva, Pinterest and Luma rolled out creative tools. Google also expanded AI-content checks, and Abu Dhabi teamed with OpenAI to make government rules machine-readable.

European landlord Heimstaden is weaving OpenAI tools into leasing, procurement and property analysis, with staff retaining oversight. It says AI cut virtual staging costs and accelerated a major environmental-risk assessment.

Formerly TesterArmy, e2e has raised $8 million to build open-source tools for testing software made by coding agents. Its web and mobile framework has drawn 1.5 million npm downloads in its first month.

Codex’s growth pushed OpenAI to build Pando, a Git-compatible backend now serving its monorepo and about 1 petabyte of Git data daily. OpenAI says it cut median push latency by 80% and improved merge reliability.

Anthropic plans to exceed 3.5 gigawatts of computing capacity by year-end 2026, underscoring how power, chips and data centres are becoming as crucial to the AI race as better models. Its long-term commitments reportedly reach $895 billion, even as losses raise questions ahead of a potential IPO.

BBVA unveiled a framework to govern AI agents as it prepares for wider deployment in 2027, but hasn’t disclosed how it will test or approve them. Pilots show promising results, though key measures of accuracy and quality remain unclear.

OpenAI and Anthropic are rolling out cheaper AI models to attract more customers, putting pressure on premium offerings. OpenAI cut API prices by 50%; Anthropic says typical workloads can cost up to 40% less.