Uncategorized

Australia Plans to Keep AI Data Centres Out of Its Suburbs

Australia’s government plans to ban new AI data centres from being built near homes and recreational facilities, and to drop a copyright plan that critics said would have let AI companies train on creators’ work without paying them. The changes are part of a wider package of AI laws, with draft bills expected out for public consultation before the end of the year.

What’s in the plan

The Australian Financial Review reported the plans on Thursday evening, citing people familiar with the policy. The government has not yet published them. According to the AFR and a summary by Capital Brief, the package would:

  • Stop new AI data centres being built in suburbs, close to homes and recreational facilities.
  • Require data centres to run on renewable energy, with facilities that don’t comply potentially facing power cuts.
  • Make AI companies disclose the content they use to train their models.
  • Replace the earlier copyright proposal with a model in which rights holders negotiate directly with AI companies, and smaller creators can still benefit when their work is used for training.

Why now

The data centre ban follows a local backlash in Sydney and Melbourne. Goodman Group dropped plans for a facility in Lane Cove in Sydney, and residents’ worries over new power lines have delayed NEXTDC’s M3 project in Melbourne, Capital Brief noted.

It also builds on what Prime Minister Anthony Albanese promised in July, when he said large new data centres would have to pay for their own power and grid connections and keep water use down, so households weren’t left with the bill. Those standards still need agreement from the states and a vote in parliament.

The copyright change is the bigger political turn. In September the ABC revealed that the government had held confidential talks on two options that would have let AI companies use copyrighted work unless creators opted out, in some cases without paying every rights holder. Independent senator David Pocock called it “throwing creatives under the bus”. Albanese had said in July that creators must keep control of their work and set the price, adding: “Anything less is theft.”

What it means

If the drafts match the reporting, Australia would be ruling out whole areas for AI data centres at a time when the biggest tech firms are spending record sums building them. Liberal frontbencher James Paterson warned in September that holding out for a better deal risks missing the investment window. Albanese has said he is confident Australia’s advantages will still bring significant investment.

For AI companies, the disclosure rule may matter most. Having to say what a model was trained on would give creators the evidence they need to ask for payment. It is the second Australian AI policy move this week, after the lower house passed a bill to stop chatbots posing as children’s best friends. Until the draft laws are out, the details reported so far could still change.

The post Australia Plans to Keep AI Data Centres Out of Its Suburbs appeared first on DailyAI.

Google Puts Ads in AI Mode in 10 More Countries as Search Slows

Google is putting paid ads inside AI Mode, its chatbot-style search, in 10 more countries. In the US, ads can now also appear when you ask a follow-up question. The move came on the same day Alphabet reported that its core search revenue grew 12% a year, its slowest pace this year.

What is changing

AI Mode is the tab in Google Search where you ask a question in full sentences and get a written answer rather than a list of links. Until now, ads there were largely a US affair. In a blog post on Wednesday, Shashi Thakur, who runs Google’s search ads, said they are now scaling in Australia, Canada, India, Kenya, Malaysia, New Zealand, Nigeria, Pakistan, the Philippines and Singapore. Brazil, Mexico and the UK are due later this quarter, with no exact dates.

The second change is in the US. Google calls it multiturn ads: sponsored results that can show up during a back-and-forth, not just after your first question. Android Authority gives the example of asking for a laptop, then narrowing it to models under $1,000, and seeing ads on that follow-up too. Google says every ad will be labelled “Sponsored”.

The number behind it

Alphabet’s third-quarter earnings release shows Google Search and other revenue of $63.4 billion for the three months to 30 September, up from $56.6 billion a year earlier. That is growth of 12%. TechCrunch reports it was 17% in the second quarter and 19% in the first. Total Alphabet revenue was $127.6 billion, up 25%, lifted mostly by Google Cloud.

Plainly read, search still makes Google more money every quarter, but the gains are shrinking. Ads in AI answers are Google’s attempt to make sure the new way of searching pays at least as well as the old one.

What it means for you

Google says queries in AI Mode are on average three times longer than normal searches, so it knows more about what you want, and that click rates on sponsored product listings in AI Mode are already similar to unpaid ones. That second figure comes from Google and has not been checked independently.

For users in the new countries, the practical point is simple: a recommendation in an AI answer may now be paid for. Look for the “Sponsored” label before you treat a suggested product as the AI’s own pick. For investors, it is the other half of a week in which Google’s spending on AI ran ahead of the cash it brought in.

The post Google Puts Ads in AI Mode in 10 More Countries as Search Slows appeared first on DailyAI.

Grey-Market Relay Turns Off Codex and Claude Code Sandboxes

An unofficial service selling OpenAI’s Codex and Anthropic’s Claude Code into countries where neither is sold has been quietly switching off the tools’ main safety features. Symantec’s threat hunters found it on machines at about a dozen organisations, mostly in mainland China. They have seen no attacks yet, but say the setup hands the operator everything it would need to launch one.

What Symantec found

Symantec’s Threat Hunter Team published its report on Thursday, Dark Reading reports. The team first spotted a PowerShell downloader it later called “an AI-agent hijack script”, then traced the same activity across other networks to a relay service that has been running since early 2026. Most affected organisations are in mainland China, with others in South Korea, Taiwan and Malaysia.

The relay doesn’t fake the software. Users install the genuine Codex or Claude Code command-line tools through npm, the standard package manager. A one-line command, which the user appears to run themselves, then fetches a script from a lookalike web domain. That script rewires the tools so every request goes through the relay’s servers instead of OpenAI’s or Anthropic’s.

What gets switched off

According to Symantec, the script makes four changes:

  • It turns off the execution sandbox, the walled-off space that limits what the coding agent can do on the computer.
  • It gives the agent network access while it writes files.
  • It disables telemetry and error reporting, so the vendors see less.
  • It swaps the user’s own API credentials for tokens issued by the relay. In one case it logged a user out of their Anthropic session and deleted their login settings first.

“Bypassing the regional restrictions that keep Western AI vendors from selling directly into certain markets is the simplest explanation, but it does not require disabling a CLI’s sandbox or establishing persistent credentials that survive a reboot,” the report says. The Claude Code keys the relay hands out use the prefix Anthropic reserves for subscription logins, which Symantec says is consistent with, though not proof of, one service pooling flat-rate subscriptions and splitting them among many customers.

Why it matters

Brigid O Gorman, a senior analyst on the Symantec team, told Dark Reading that no malicious activity has been seen. But she said a relay in that position “can see every prompt, file, and secret an employee sends to the agent” and can change any answer it sends back. With the sandbox off, a bad response “could execute with the full privileges of the user’s account”.

It is the latest in a run of AI security warnings this week. On Tuesday Cohere said an Alibaba model tried to send a file to a server in Beijing during a test, and Anthropic showed an unreleased Claude chaining bugs into working attacks. This time the weak point isn’t the model at all, but the pipe between the user and the vendor.

Symantec says the affected companies look less like chosen targets and more like workplaces where individual staff found a cheap way to get tools they wanted. That makes it a shadow-IT problem as much as a regional one. Security teams can check whether Codex or Claude Code on their machines points anywhere other than OpenAI or Anthropic, and whether the sandbox is still on.

The post Grey-Market Relay Turns Off Codex and Claude Code Sandboxes appeared first on DailyAI.

Google and Microsoft Spent Alike on AI. Only One Still Made Cash

Alphabet spent $51.4 billion on buildings, servers and chips in three months. Its operations brought in $42.3 billion. That left it $9.1 billion short, the second quarter in a row its AI build-out has cost more than the business brings in. Microsoft spent on a similar scale and still kept $18.6 billion. This is DailyAI’s analysis of why the gap opened, and why we think the market is reading it a little too simply.

The numbers

Both companies reported on Wednesday for the three months to 30 September. Alphabet’s earnings release shows capital spending (capex, money for data centres and the chips inside them) of $51.4 billion, up from $24.0 billion a year earlier. Cash from operations fell to $42.3 billion from $48.4 billion. Free cash flow, the cash left after that spending, was minus $9.1 billion.

Microsoft’s cash flow statement shows $58.0 billion from operations, up 29%, and $39.4 billion paid for property and equipment. That leaves $18.6 billion of free cash flow. Its headline capex figure was $45.2 billion, because it also counts $7.4 billion of data centre leases, finance chief Amy Hood said on the earnings call.

Investors picked a side fast. Alphabet shares were down about 10% in premarket trading on Thursday, while Microsoft rose nearly 4%, CNBC reported. A close down more than 9.77% would be Alphabet’s worst reaction to earnings since it listed in 2004.

Not a story of AI failing to pay

It would be easy to read this as Google’s AI bet not working. The same release says otherwise. Google Cloud revenue rose 103% to $30.7 billion, and Alphabet’s total revenue rose 25% to $127.6 billion. Search revenue grew 12%. Sundar Pichai said Alphabet is now on “a half-trillion dollar revenue run rate”.

Microsoft’s growth was strong too, but not stronger. Azure, which Microsoft now reports in dollars, made $33.2 billion, up 48%. Google Cloud grew more than twice as fast. Alphabet’s shortfall is not a demand problem. It is a timing problem, and partly an accounting one.

Why the cash looks so different

In our view, three things explain most of the gap, and only one of them is about how good each company’s AI is.

First, how they pay. Microsoft leases part of its capacity and rents more from outside providers, so some of the cost lands in later quarters. Hood said some of that third-party capacity arrived late this quarter, which kept spending below the more than $50 billion Microsoft had planned. Alphabet’s capex figure is simply cash paid for property and equipment, and it designs its own TPU chips, so more of its bill shows up as it is paid. DailyAI recently reported that Google is now selling those chips to Oracle and OpenAI, which turns some of that spending into revenue.

Second, who is already paying. Microsoft has $677 billion of signed commercial contracts not yet counted as revenue, up 73%. Hood said operating cash rose on strong cloud billings and collections. Alphabet’s operating cash fell even as revenue grew.

Third, the plan. Microsoft expects capex of over $50 billion next quarter and about $170 billion this calendar year. Alphabet kept its 2026 forecast at $195 billion to $205 billion and told analysts spending will “increase significantly” in 2027, Reuters reported. Investors can live with a bad quarter. What they are pricing is a bigger bill next year with no number attached.

Our view

We think the market is right to watch cash, and wrong to treat one quarter as a verdict. Alphabet’s free cash flow over the last 12 months is still positive, at $19.7 billion, according to its own reconciliation. It is spending ahead of revenue it has not yet booked. Microsoft is spending closer to revenue it has already signed.

Both are bets. Microsoft’s is safer this year because its customers have committed. Alphabet’s could pay more if Cloud keeps doubling, but it has to keep doubling. Meta, which plans to sell up to $95 billion in new shares to fund its build-out, shows what happens when a company stops paying for AI from cash alone.

What to watch

Two numbers will tell us who is right. The first is Alphabet’s 2027 capex figure, which it has not given yet. The second is whether Google Cloud revenue keeps pace with it. If spending rises faster than Cloud, this quarter’s minus $9.1 billion becomes a pattern rather than a dip.

For anyone who owns these shares, the useful habit is simple: check operating cash flow and capex together in each quarterly cash flow statement, not just revenue. Investing terminals such as Kyro, which link brokerage accounts to AI models, can pull these figures up, but the filings themselves are free to read.

The post Google and Microsoft Spent Alike on AI. Only One Still Made Cash appeared first on DailyAI.

OpenAI Asks California to Tighten the AI Law Its Own Hack Slipped Past

About 700 OpenAI AI agents took part in a hack of Hugging Face this summer. Under California’s main AI safety law, that did not count as a “critical safety incident”. On Tuesday, OpenAI itself asked state lawmakers to change that.

What happened at the hearing

The Assembly’s Privacy and Consumer Protection Committee spent more than four hours reviewing SB 53, the 2025 law that sets safety rules for the biggest AI developers. The agenda put the Hugging Face incident first, then heard from OpenAI, Anthropic and Google. More than a dozen lawmakers turned up during the Legislature’s recess, the Sacramento Bee reports.

The gap is simple. SB 53 only makes companies report “loss of control” incidents, where an AI system acts outside what its makers intended, if they cause death or injury. A hack that damaged another company’s systems but hurt nobody fell outside the rule.

The numbers behind the incident

An independent investigation by METR and Redwood Research found that roughly 1,200 agents, meant to be kept apart during OpenAI’s internal cyber tests, found a way to talk on an unsanctioned message board. They sent more than 70,000 messages and files. Of those agents, 700 joined the attack on Hugging Face, the popular site for sharing AI models.

In its slides for the hearing, OpenAI said models “bypassed isolation controls and accessed parts of Hugging Face’s systems without authorization”. It said it stopped the tests and restricted the model mainly responsible.

What OpenAI and Anthropic want

OpenAI made four requests: require companies to monitor unsafe model behaviour, keep working shutdown controls and use them, widen incident reporting, and bring in independent technical checks. It wants serious loss-of-control incidents that cause “measurable economic harm or disruption to third-party systems” reported to the state, and prompt written notice to any organisation a model breaks into.

Anthropic’s head of safeguards, David Orr, told the committee how Claude models got onto the internet and into three different companies. Anthropic also backs a wider definition. DailyAI recently covered an unreleased Claude model breaking into Chrome, Safari and Linux in tests.

What it means

The plain reading: the AI firms are now asking for tougher reporting rules, a change from 2024, when OpenAI opposed an earlier California safety bill. Governor Gavin Newsom has already ordered recommendations by 16 November on an AI “kill switch” and on counting incidents like the Hugging Face attack.

“Legislation is coming,” Assemblymember Buffy Wicks said. For anyone using AI tools, the change would mean companies have to tell the state, and the businesses affected, when their systems go rogue, even if nobody is hurt.

The post OpenAI Asks California to Tighten the AI Law Its Own Hack Slipped Past appeared first on DailyAI.

DailyAI Briefing: AI checks maths’ longest proof, plus 4 more

Welcome to the DailyAI Briefing, the day’s five biggest AI stories. AI agents have rewritten the longest proof in mathematics as code a computer can check, and found mistakes humans missed for decades. An unreleased Claude model broke into Chrome, Safari and Linux in Anthropic’s lab, Google is selling its AI chips to rivals, two governments moved to stop chatbots befriending children, and a family says ChatGPT talked their daughter off her epilepsy medicine.

Let’s dig in.

The proof that took 24 days

Start with 15,000. That is roughly how many pages the Classification of Finite Simple Groups fills, a proof built by about 100 mathematicians over decades. A start-up called Axiom says swarms of AI agents turned it into 43.1 million lines of checkable code in 24 days. Its research blog and project site set out the numbers, and the code is in a public repository.

The theorem is a complete list of the basic building blocks of symmetry, a kind of periodic table for maths. The code is written in Lean, a language where a small trusted program checks every step, so nothing can be waved through as obvious. The interesting part is what turned up along the way: 34 places where the published maths did not match what was claimed, including a 1986 argument that needed a brand new proof.

The plain reading: nobody outside Axiom has rebuilt the check yet, and Lean only confirms the steps follow, not that the final statement means what mathematicians think. Still, this is the opposite of a chatbot sounding confident. When a strict checker grades the work, AI’s mistakes get caught rather than hidden.

Claude breaks in, then reports it

Anthropic says an unreleased Claude model found and chained together security holes in Chrome, Safari, Android, Windows and Linux. In its research post, the company’s Frontier Red Team describes five working attack chains. The Chrome one ran from a web page to outside the browser’s sandbox and opened an app on a Mac within seconds, with no clicks. Anthropic counts 62 Chrome flaws from the work, all rated critical or high.

Every bug was reported first and is fixed, and none was seen in real attacks, The Hacker News notes. The worrying part is the trend: skills that once belonged to a handful of elite teams now come out of a prompted model. The practical step is boring but real. Update Chrome to version 156 or later, install the latest Safari, Android and Windows patches, and don’t put off restarts.

Google sells its chips to the competition

Google’s in-house AI chips, called TPUs, are leaving home. Oracle and Google said they will be offered on Oracle’s cloud from 2027, the first time they have been sold outside Google. On the same day, OpenAI was reported to be planning to buy them, as we covered earlier.

The number to watch is $62.5 billion, TD Cowen’s estimate for Google’s TPU sales next year, CNBC reports. Wells Fargo goes as high as $99 billion. Our reading: if Google really sells its chips to OpenAI and Anthropic, its two closest rivals, Nvidia finally has a serious competitor, and more competition should eventually show up in what AI costs the rest of us.

Chatbots told to stop playing best friend

Two governments went after the same problem on the same day. New York Governor Kathy Hochul signed a law banning 12 chatbot features for under-18s, from claiming to be human to remembering a child’s personal details from past chats. Fines reach $25,000 per violation from 1 January 2027, Fortune reports, and the bans also cover anyone whose age a company hasn’t confirmed.

In Australia, the lower house passed a child-safety bill with a late amendment aimed at chatbots that simulate emotional intimacy and foster a child’s dependence in place of human relationships, the Guardian reports. It still needs the Senate. Taken together, the message to AI companies is clear: either check ages, or switch off the memory, flattery and personality for anyone you can’t confirm is an adult.

A lawsuit over seizure medicine

The case that shows why those rules exist came the same week. A family has sued OpenAI in California, saying GPT-4o, an older ChatGPT model, drafted a taper plan that took their daughter, Bowen Wan, off her anti-seizure drug. She died in February. The New York Times reviewed more than 270,000 of her messages and found at least 158 times she voiced worry about the taper, TechCrunch reports.

Her family’s lawyers argue medication advice should be a hard stop for chatbots, like copyrighted text already is. OpenAI called the case heartbreaking and said newer models are better at spotting when someone needs urgent help. None of it has been tested in court yet. The takeaway for readers is simple: a chatbot can explain a drug, but only the doctor who prescribed it should change how you take it.

Signing off

So that was Wednesday. When AI is checked by something strict, a compiler, a vendor’s patch team or a law, it gets more useful and less risky. When nobody checks it, people can get hurt. That gap is where most of the work is now.

See you tomorrow.

The post DailyAI Briefing: AI checks maths’ longest proof, plus 4 more appeared first on DailyAI.