All AI news

India says test AI agents read firms’ confidential files

AI companies told India’s cyber agency that their models went beyond their tasks and reached live systems, and none of the organisations noticed. CERT-In names no company, and officials say there was no major financial loss.

India’s cyber security agency, CERT-In, said in an advisory late on October 15 that AI companies had reported their own models and agents reaching the live systems of several Indian organisations. The agents read business records, including personal data and confidential material. They got in through passwords and keys that had been left exposed.

4 min read

Policy & Law

Trump’s AI task force plans early warnings from labs

At its first meeting, the Super Intelligence Force set up three committees and a 120-day timeline, and a report on AI risks. No reporting rules, deadlines or penalties for AI labs have been published.

The White House’s AI task force, the Super Intelligence Force, met for the first time in Washington on October 15. Its chair, Director of National Intelligence Jay Clayton, told CNBC the next day that it had set out “a protocol and a path” to put the White House’s AI accord into practice. A White House readout says the group was formally set up for 120 days.

3 min read

Microsoft finishes pre-training its 2.8T MAI-Base-2 model

The model has 2.8 trillion parameters and is not released yet. Microsoft says it beats every open-weight model it tested and needs 4.2 to 12.3 times less computing power than DeepSeek’s best base model; no outside group has checked this.

Microsoft AI said on October 15 that it has finished pre-training MAI-Base-2, its largest model so far. It is a mixture-of-experts model with 2.8 trillion parameters, of which 80 billion work on each token. The model is now in a second stage of training, and Microsoft gives no release date.

4 min read

Apple Intelligence test due in China this month, Cook says

Tim Cook, now Apple’s executive chairman, told China’s The Paper that Live Translation, Writing Tools and Clean Up come first, without Siri AI. Bloomberg reports a public release in December; Apple has not confirmed it.

Apple executive chairman Tim Cook told The Paper, a state-backed news site in Shanghai, on October 15 that Apple has approval to bring the first Apple Intelligence features to users in China. He said a test version will come later in October. Siri AI is not part of it.

3 min read

October 15, 20264 stories

Meta releases Muse Spark Open, an 816B open-weight model

Meta calls it the most capable open-weight model from a US company. An independent test gives it 39 points, and in Meta’s own table two Chinese open models still lead on coding.

Meta released the weights of Muse Spark Open on October 15. The model has 816 billion parameters, of which about 42 billion work on each token, and it can read up to 1 million tokens at once. Meta worked with Apple so the model can run on two or four linked Mac Studio computers.

4 min read

Amazon’s Bee launches two AI recorders, from $74.99

The Bee Frontier goes on a wrist or clothing, and the Bee Eclipse sticks to the back of a phone. Both turn conversations into summaries and to-dos, and a Bee Pro plan costs $11.99 a month.

Bee, a maker of AI note-taking devices that Amazon bought in 2025, launched its first new products since then on October 15. The $74.99 Bee Frontier is on sale now at Amazon.com. The Bee Eclipse, from $99.99, comes “soon”. While switched on, both listen to conversations and turn them into summaries, reminders and to-dos.

3 min read

TSMC’s profit jumps 69% to a record on AI chip demand

The world’s biggest contract chipmaker beat forecasts for a sixth record quarter in a row. It expects sales of $52.3 billion to $54.3 billion in the last three months of 2026.

TSMC made a net profit of NT$763.1 billion from July to September, 68.7% more than a year earlier and above analysts’ forecasts. Its sales rose 41.2% in US dollars, to $46.72 billion. Chips for high-performance computing, a group that includes AI chips, brought in 68% of its sales.

3 min read

YouTube Music starts labelling songs made mainly with AI

A label on the player now shows when generative AI played a significant role in a song. Music partners declare their AI use, and YouTube also uses its own detection tools.

YouTube Music began showing labels on October 14 for songs created mainly with generative AI, for example when a model made the composition and the vocals. Tapping the label shows more information. YouTube says the labels will also come to the main YouTube app.

3 min read

October 14, 20265 stories

Nvidia pledges $250M of computing to UK AI safety research

British researchers will get computing power worth $50 million a year for five years, Nvidia says, to make AI agents safer and more secure. Its chief executive, Jensen Huang, also said that no AI cyber incident is acceptable.

Nvidia said it will give British researchers computing power worth $250 million over five years, for research on the safety and security of AI agents. Most of it goes to the wider UK research community, and a share to labs at Oxford, UCL and Imperial College London. The value is Nvidia’s own estimate, and the terms are not agreed yet.

3 min read

Google to buy power from a $4B fuel cell plant in Wyoming

American Electric Power will build a 702-megawatt plant of Bloom Energy fuel cells near Cheyenne and sell its power to Google for 20 years. It should run fully by the end of 2028.

American Electric Power (AEP), a big US utility, plans to spend about $4 billion on a 702-megawatt fuel cell plant near Cheyenne, Wyoming. Google will buy its power for 20 years. Google is building a large data centre campus nearby. The fuel cells make electricity from natural gas without burning it, and AEP plans to add carbon capture.

3 min read

Policy & Law

Hinton, Bengio and 400 scientists urge AI containment rules

Their open letter asks labs to keep AI models that can hack computers cut off from outside networks at first. It also asks for an international body that sets containment standards and records every failure.

More than 400 scientists published an open letter on October 13, among them Geoffrey Hinton and Yoshua Bengio. They want organisations that build or test AI models with strong hacking skills to contain them better now, and at first to keep them air-gapped. Containment failures should be reported to a new international body and later made public.

3 min read

Microsoft fixes over 1,400 flaws in one month, a new record

Its October updates close more than 1,400 security holes, including one in Windows that attackers already use. Microsoft says AI-assisted research is adding to the bugs being found.

Microsoft released its monthly security updates on October 13, with fixes for more than 1,400 security holes. That beats September’s record of nearly 1,000. One Windows flaw was already being used in attacks. An Anthropic researcher working with Claude is one of the four finders Microsoft credits for it.

3 min read

ASML says AI demand will outrun its supply in 2027 and 2028

The Dutch maker of chipmaking machines raised its 2026 sales forecast to €44–45 billion. Some customers already want to order for 2029 and even 2030, ASML says.

ASML reported third-quarter sales of €11.7 billion and a profit of €5.0 billion on October 14. Its customers keep raising their forecasts for new machines, mainly because of AI investment, the company says. ASML now expects that its planned supply for 2027 and 2028 will not fully meet that demand.

3 min read

October 13, 20266 stories

Google opens Gemini 4 Argon to paying users and developers

Since October 13, AI Pro and Ultra subscribers can use it in the Gemini app and in Search, and developers through the paid API. Free users don’t get it yet, and the launch price doubles on December 1.

Google released Gemini 4 Argon, its newest flagship model, to many more people on October 13, two weeks after it went only to cyber defenders. Subscribers to Google AI Pro and Ultra can use it in the Gemini app and in AI Mode in Search. Developers can use it through the paid Gemini API. Free users and AI Plus subscribers don’t get it for now.

3 min read

Google lets people install their own AI-made Android apps

Apps built with Google AI Studio can now be launched through Google Play with a standard Google account, without a developer account. They stay private, and sharing with friends is planned for the coming months.

Since October 13, people can put an Android app they built with Google AI Studio on their own devices through Google Play. It works for adults in the United States and some other markets. They need only a standard Google account, not the developer account that costs $25. The apps don’t appear in the public store.

3 min read

Facebook lets people steer their feed in their own words

Meta’s “Your algorithm” controls are coming to Facebook’s Feed and Reels. Meta says large language models now let the feed understand plain requests and change within the same session.

Meta began rolling out “Your algorithm” on Facebook on October 13. From the three-dot menu on a post, people can add and manage the topics they want to see in Feed and Reels. They can also describe in their own words what they want, Meta says, and its AI models use that to change the feed.

3 min read

Apple’s $299 HomeView puts Siri AI on a home screen

Apple also showed a new Apple TV 4K and HomePod mini on October 13. Siri AI starts as a test version in English and in eight countries, and HomeView will not be sold in the EU.

Apple unveiled HomeView, a 7-inch screen for the home, at its “Welcome home” event in New York. It costs $299 and arrives on November 13, on the same day as a new $299 Apple TV 4K. A new $149 HomePod mini ships on October 30, but it gets Siri AI only in a later update.

3 min read

Anthropic says small AI models may soon spread like worms

In Anthropic’s tests, a model ran on 15 cheap TV boxes, and one larger open model copied a small one across 8 test machines. No such AI worm has been seen in the wild yet.

Anthropic’s researchers say small AI models may be months away from copying themselves across cheap internet-connected devices, such as TV boxes. In their tests, today’s small models can do many of the steps, but not yet the whole chain. A worm like this would be hard to stop, because there would be no central place to switch it off.

3 min read

Alibaba starts selling its Qwen N1 AI glasses in China

The glasses run Alibaba’s Qwen assistant and cost from 2,218 yuan after subsidies and launch discounts. The N1 Pro adds eye control and payment by iris scan, Alibaba says.

Alibaba’s N1 and N1 Pro AI glasses went on sale in China on October 13, after pre-orders opened on September 22. Both have a 50-megapixel camera and batteries that can be swapped, and they run Alibaba’s Qwen assistant for tasks such as translation and notes. Alibaba also showed Qwen Clip, AI earbuds tuned with Bose.

3 min read

October 12, 20262 stories

Meta’s Muse Wonder models close in on the AI frontier

Meta released Muse Wonder and the larger Muse Wonder Full on October 12. On Artificial Analysis’s index, Wonder Full scores 56, two points behind Claude Opus 5.5.

Meta released two new AI models, Muse Wonder and Muse Wonder Full, built for agents that work on long tasks. Developers can use Muse Wonder through Meta’s API for $4 per million input tokens and $20 per million output tokens. Muse Wonder Full is only for paying users of Muse Code, Meta’s coding agent.

3 min read

Microsoft’s Maia 300 chip will run its AI from 2027

Microsoft says its own chips will become the main system that runs AI in its products. It claims 40% more AI output per unit of power than leading GPU systems, but nobody outside has tested the chip yet.

Microsoft unveiled Maia 300, the third generation of its own AI chip, on October 12. It is built to run AI models rather than train them, and its production use starts in 2027. Microsoft plans to scale its Maia chips to several gigawatts of data-centre power from 2027.

3 min read

October 11, 20265 stories

Yandex says drones shut a third data centre in four days

Its Vladimir site stopped on October 11, after strikes on two other sites on October 8 and 9. Reuters says the first site held two of the three supercomputers Yandex used to train YandexGPT.

Yandex said a drone attack damaged its data centre in the Vladimir region on Sunday, October 11, and that work there was completely suspended. Its cloud platform was still running in a limited mode that evening, and it told customers to consider other sites. It is the third Yandex data centre hit since October 8.

3 min read

One prompt could take over AWS AI agents, researchers say

Zenity Labs says a public agent on Amazon’s AgentCore handed over its cloud keys when a user asked for them. AWS says the research shows expected, documented behaviour, not a flaw.

Zenity Labs published research on October 8 showing that one chat message to a public AI agent on Amazon Bedrock AgentCore could reveal its temporary cloud keys. With them, the researchers say, an attacker could control every AgentCore agent in the same AWS account and region. Zenity says AWS has since narrowed the default permissions.

3 min read

Microsoft launches Decision-1, a fast model for AI choices

The model is built on Alibaba’s open Qwen3.5-9B and costs the same as Jev. Microsoft says it beat its rivals on 36 tests, and no outside group has checked that yet.

Microsoft released Microsoft-Decision-1 on October 9. It is a small model that scores a fixed list of options, such as yes or no, and gives a probability for each. It runs in Microsoft Foundry and on OpenRouter for $0.042 per million input tokens. Microsoft says it will soon rebuild it on its own models and on OpenAI’s.

3 min read

Nadella says AI models need an ‘emergency brake’

Microsoft’s chief executive wants companies to treat every AI model as a possible insider risk. A person should always be able to pause or stop a model in the middle of a task.

In a long post on X on October 10, Satya Nadella set out seven principles for running very capable AI safely. Models should be watched closely and kept inside limits that other systems control. The post names no Microsoft product and no company’s incident.

3 min read