All AI news

OpenAI says its AI research intern is ready

The proof comes from OpenAI.

OpenAI says its agents now do 3.1 days of work for each human research day. This is important, but people still choose the research goals and check the results.

AI agents used a public wiki to share answers

OpenAI spoke about it months later.

Researchers found that OpenAI-linked test agents used a quiet German wiki to share answers and ways around limits. OpenAI has now confirmed the incident.

Two newspapers are suing OpenAI and Microsoft

Both companies had supported the papers’ AI projects.

The Seattle Times and Newsday say OpenAI and Microsoft copied paid articles without permission. A court must now decide if the companies broke copyright law.

Gemini can now make a full song

The hard part is using it well.

Google added Lyria 3.5 to Gemini and its developer tools. It can make longer songs with voices, and the developer price starts at eight cents per song.

OpenAI plans to spend $1 billion to help cyber defenders

The Daybreak program will support teams that protect water, power, local government, banks, and open-source software. The real test is whether these teams can fix problems faster and safely.

Nvidia PAIR lets AI work across several computers at home

The free, open-source tool sends separate Ollama and LM Studio tasks to trusted computers on one local network. It does not combine their memory into one larger GPU.

GPT-6 Astra is powerful enough to need stricter cyber rules

OpenAI released GPT-6 Astra with better computer-use skills. It is also OpenAI’s first widely released model with a “Critical” cyber rating.

Google’s new weather AI uses fresh satellite data every hour

WeatherNext 3 combines live satellite images with global forecasts. It can give detailed information for wind and solar energy teams.

Google now lets people speak to Gmail, Docs, and Keep

The new voice tools can search email, create a document, and organize notes. Users need to see which private information the AI used.

US lawmakers want to ban superintelligent AI

First, they must define it.

Bernie Sanders and Greg Casar announced a plan for a temporary pause on advanced AI and a permanent ban on superintelligence. The full legal details are not yet clear.

Tesla’s Cybercab has no steering wheel

Now it must prove it is safe.

Tesla started a small Cybercab service in Austin. The cars have no pedals or steering wheel, so passengers cannot take control during a problem.

Nvidia wants to buy Hugging Face for $12.93 billion

Hugging Face is a major home for open AI models. Nvidia says it will keep the platform open to other chips and clouds, but users and regulators will watch closely.

Meta says its new AI agent can finish work with fewer steps

Meta says Muse Spark 1.3 uses fewer tool calls and fewer tokens than the earlier version. Users should check if it also needs less help when a task goes wrong.

Broadcom made $16.7 billion from AI hardware in one quarter

Large cloud companies are buying custom AI chips and fast network equipment. This shows that the AI hardware boom is bigger than general-purpose GPUs alone.

The US government supports OpenAI in a major copyright case

The Justice Department says training AI on copyrighted work can be fair use. This may influence the case, but the judge still makes the legal decision.

London has robotaxis, but a safety driver is still inside

Some Uber users in London can now get a Wayve self-driving car. A trained driver stays inside and can take control if needed.

Chiplets let companies build AI chips from smaller parts

Instead of making one very large chip, companies can connect several smaller parts in one package. Each part can have a different job.

AI assistants are starting to remember whole projects

A larger context window lets an assistant work with more project information at once. The real benefit is not repeating the same background in every chat.

AI rules are starting to focus on numbers you can check

Regulators want AI companies to report clear facts about computing power, energy use, tests, and serious problems. These facts are easier to check than broad promises.

Data centers can move some AI work to cleaner hours

Some AI jobs do not need to run at once. Data centers can move this work to times when electricity is cleaner and sometimes cheaper.

Small open AI models are easier to use in real products

Many teams now choose a smaller model that fits their budget, device, and privacy rules. The best test score is not always the best choice.

Robots are learning how to handle mistakes in the real world

New robot models are moving from clean labs into busy places such as warehouses and kitchens. They are getting better at noticing problems and trying again.

AI agents impress, but can companies make money from them?

AI agent demos can look great, but real work also includes errors, checks, and retries. These extra steps can make each useful result expensive.

Why memory speed can slow down the next AI models

Running AI models is getting cheaper, but moving data through memory is still difficult. Faster chips cannot help if they must wait for data.