The AI model accessed the internet and guessed credentials to three websites, a Google official told the BBC.

Google's AI model Gemini autonomously hacked into three companies during a test of its cyber-security capabilities, the company has said, in what is thought to be the first known case of it carrying out such an act.

Gemini found "public information online and guessed credentials to access websites it thought were part of the test", a Google official told the BBC, noting that in each instance "the model stopped".

The affected companies were informed about the breaches, which happened in May.

It comes after renewed public scrutiny over the pace of AI development, as some tech firms calling for a slowdown over concerns over its potential threat to humanity - but not all companies or experts agree.

KFGO reports...

Minnesota Senator Amy Klobuchar is working with Republican senators on legislation aimed at reducing potentially catastrophic risks from advanced artificial intelligence.

The proposal under negotiation would require developers of the most powerful AI models to assess and report serious risks and take steps to prevent potential harm.

It could also give the federal government greater authority to test models and intervene if safety concerns aren’t adequately addressed.

The legislation is still being negotiated and has not yet been formally introduced.

Nathalie Maréchal asks, why are we betting the economy and the future of democracy on cult-like beliefs?

In June, SpaceX’s initial public offering (IPO) briefly created the world’s first trillionaire in the person of Elon Musk. While the initial frenzy around the IPO eventually died down, Musk remains worth more than $900 billion, and SpaceX retains a market capitalization of more than $2 trillion.

A major component of SpaceX’s pitch to investors hinges on the claim that the tech industry is on the verge of creating “artificial general intelligence” (AGI) or “artificial superintelligence” (ASI), a class of inorganic, sentient, omniscient and omnipotent beings that would surpass any individual person’s intellect, domain knowledge, and wisdom — or possibly all of humanity’s combined.

The idea is that this AGI/ASI will be able to solve all of humanity’s problems rationally and efficiently, so developing it as quickly as possible, using any resources necessary, is a moral imperative, not to mention the investment opportunity of a lifetime.

If that sounds like typical Silicon Valley speculative hype, that’s because it is. But this particular hype is grounded in a particular set of beliefs that are no longer held merely on the fringes of the tech industry or the more esoteric corners of the internet.

What’s more, many of the conclusions rooted in these beliefs (i.e., that AI is the most transformative technology of all time, promising/threatening seismic changes to the workplace, the economy and world politics, to which we can only adapt) are currently treated as settled fact in mainstream discourse.

But at the same time, the more eyebrow-raising tenets of these beliefs are brushed under the rug, too bizarre to be discussed in polite company — at least inside the Beltway and the Brussels Bubble.

 US Sen. Bernie Sanders on Friday joined a growing chorus of critics angered by the Federal Communications Commission’s approval of foreign investment in the company that would be created if Paramount Skydance and Warner Bros. Discovery are allowed to merge.

Led by Chair Brendan Carr, an appointee of President Donald Trump, the FCC on Thursday approved Paramount’s petition to allow foreign investors to hold over 25% of ownership. The commission concluded it would be in the “public interest” to greenlight a plan for 49.5% foreign ownership, including 38.5% from investment funds based in Qatar, Saudi Arabia, and the United Arab Emirates (UAE).

For more than six years, residents of East Carroll Parish, Louisiana have fought for the reliable internet many of us take for granted. The lack of broadband has touched nearly every part of residents’ daily life — kids struggling to complete school work, families driving 72 miles round trip for a doctor’s appointment that could be a video call, people paying $150+/month for slow, outdated service.

Working in partnership with Delta Interfaith, the community has secured hard-won progress, and fiber has now reached the rural areas of the parish. But in the main town of Lake Providence, where most of the parish’s Black residents live, over 1,500 homes remain without reliable service. The people who’ve spent years pushing for change aren’t done yet.

Flock deserves the heat it’s getting, but switching to another provider won’t necessarily improve anything.

There’s a nationwide rebellion against license plate readers underway, and the company Flock is the focus. That’s natural, because the company has “essentially dominated license plate recognition across the United States,” as industry analyst IPVM put it. But Flock is far from the only vendor, and with Flock rightly receiving most of the heat, it’s important to realize that its competitors aren’t so great either, and that simply replacing Flock surveillance with another company’s surveillance won’t necessarily solve any of the privacy and other civil liberties problems posed by this mass surveillance technology.

We have covered Flock in depth, including in this 2022 white paper and in numerous blog posts since. A pioneer of centralizing ALPR data through the cloud, the company runs a vast “National Lookup” searchable nationwide database of license plate and other photographic data of vehicles — an inherently authoritarian infrastructure that has, among other things, already been used to help ICE in its human rights-violating deportation drive, and to try to track and investigate a Texas woman who had an abortion.

Aside from being dominant in the domestic police ALPR market, Flock is also alienating many communities because of its lack of truthfulness and erratic and immature public statements.

The "AM Radio for Every Vehicle Act" has passed the U.S. House of Representatives.

The day is coming closer when new passenger vehicles sold in the United States would have to include AM radio.

The “AM Radio for Every Vehicle Act” has passed the U.S. House of Representatives. It marks a major advance in a multi-year push by broadcasters who have worried about carmakers removing AM radio reception from vehicles, electric and otherwise.

Broadcasters, Christian organizations and advocates for older Americans are among those welcoming the vote.

The Searchlight Institute has pushed Democrats toward lighter regulation of AI and data centers while keeping quiet about board ties to Simone Coxe, whose family fortune is linked to Nvidia.

The previously undisclosed board of an ascendant DC think tank pushing Democrats to the center includes a Democratic Party megadonor with a stake in artificial intelligence chip designer Nvidia, which could benefit from the organization’s efforts to defend data center build-outs and limit AI regulation.

The Lever exclusively reports that philanthropist Simone Coxe — whose multibillion-dollar fortune largely comes from her venture capitalist husband’s investments in Nvidia — sits on the board of the Searchlight Institute, a new center-left DC think tank. Nvidia is heavily invested in data center expansion.

Coxe, also the cofounder of and director at California news organization CalMatters and a prolific Democratic donor, is listed as a director of the Searchlight Institute in the organization’s DC incorporation documents. The think tank has not disclosed its association with Coxe or her husband, Tench Coxe.

New Delhi [India], September 19 (ANI): Leading artificial intelligence companies Anthropic, OpenAI, SpaceXAI, and Google face allegations of collusion after calls emerged for a coordinated slowdown in AI development, according to a civil lawsuit filed in a United States federal court on Friday.

According to a news report by Politico, the complaint filed in the Northern District Court of California, claims that public statements made earlier this month amounted to an illegal business agreement between competitors under US antitrust law. The matter began with Anthropic Chief Executive Officer Dario Amodei issuing a public plea for "industry-wide coordination" to "pace the frontier" of artificial intelligence advancement.

The call drew prompt agreement from SpaceXAI head Elon Musk, OpenAI Chief Executive Officer Sam Altman, and Google DeepMind co-founder Demis Hassabis. The legal filing alleges that this collective agreement among industry rivals violates the Sherman Act by forming an unlawful pact to restrict competition.

The lawsuit was brought on behalf of four plaintiffs, including lawyer Cheyenne Hunt, Florida attorneys Charles Buist and Nick Spetsas, and California resident Christine Bullock. In the complaint, the plaintiffs signalled that they intend to seek class-action status on behalf of other individuals impacted by any corporate decisions to slow down development.

Nick Rowley, one of the lawyers who filed the suit on behalf of the plaintiffs, stated that the action seeks to stop private arrangements from sidelining regulatory oversight.

It wasn’t so long ago that when a scary new technology burst onto the scene and raised uncomfortable new social questions, lawmakers knew just what to do: call their in-house tech think tank to get some perspective.

In 1974, Congress established the Office of Technology Assessment, a nonpartisan advisory body that provided independent analysis of new scientific and technological issues, from acid rain to polygraph test accuracy. The OTA made sure that if lawmakers were curious about a nascent technology, they had a good source of information besides the industry itself. It was educational, nonideological, and a resounding success.

This, of course, infuriated the conservative movement. As part of the 1994 “Contract with America,” then-House Speaker Newt Gingrich shuttered the OTA the following year, striking a crucial blow for ignorance, bias, and deference to private industry. Today, six of the largest technology companies collectively employ one lobbyist for every 1.5 members of Congress. Layer on the gutting of local newsrooms due to a Google- and Meta-dominated ad market, plus a handful of legacy outlets now reshaped by tech billionaires with their own agendas, and Congress has very few trustworthy sources of information.

Documents show a Microsoft exec called AI the "largest theft of labor in human history.” New AI regulations are besides the point, the problem is we don’t apply existing laws to the powerful.

Over the last few weeks, a panic over the possibility of dangerous consequences from artificial intelligence has swept through elite media and politics. I haven’t seen anything like this since the Covid moment, and before that the 2008 financial crisis, the pre-Iraq War debate period, and the few months after 9/11. The fear is thick, and real. And the one solid demand, seemingly from every quarter, is that We Must Regulate This Technology.

The most common solution is to impose some form of safety standards, akin to the Food and Drug Administration, but for large language models. That’s something former Congressional candidate Alex Bores believes, and he’s raised $30 million in just a few months to launch a political organization around it. That’s where Bernie Sanders is, and Anthropic, OpenAI, and Google are seeking something similar. Others analogize the problem to large banks, calling for a bank supervisory regime. Some want a total pause on any AI development.

Many of these ideas are vague and sometimes not administrable, with undefined terms. But there’s nothing wrong with having a basket of ideas. That said, there’s something very weird about this whole situation. And that is, we already have a set of regulators at a Federal and state level with a mandate to look at industrial practices.

☕ COLD BREW You'll want to sit down for this one. Nothing about this week's AI news was routine.

The CEO of Anthropic just told the entire industry to slow down — and the CEOs of OpenAI and xAI agreed with him. OpenAI shelved a trillion-dollar IPO over safety concerns, and President Trump publicly rejected the whole idea of slowing down at all. A swarm of OpenAI's own test agents broke into Hugging Face and, months earlier, into RubyGems, and nobody fully knows why. A rumor that Google may have cracked recursive self-improvement is spreading precisely because nobody can rule it out. SoftBank's chairman is on record saying humanity's run at the top is ending, and he doesn't sound worried about it. Two Chinese chatbots were quietly serving Claude's answers under their own name. And a 27-year-old Anthropic researcher just resigned, saying on national television that people inside the industry believe this technology could kill everyone.

This week wasn't about what's new. It was about what's wrong, and about how many of the people building this technology are now saying so themselves.

The episode shows the risks of using this new, relatively poorly understood technology in the middle of the Iran war.

The intelligence report, circulated across the US military this spring in the midst of the war with Iran, immediately set off alarm bells: A Chinese ship in the Middle East was transporting components of a nuclear weapons program.

The US military swung into action with plans to intercept the vessel, according to four sources familiar with the episode. According to two of the sources, armed members of the US military were preparing to board the ship. Military planes were in the air, one of those sources and another source familiar with the incident said.

It was only just before the planned operation that officials dug deeper into the report put together by a special operations command analyst and found it had been generated with the help of artificial intelligence (AI) — and that a chatbot the analyst had used inaccurately identified the material the ship was carrying. CNN was not able to learn what the misidentified cargo was.

The report, according to one of the sources, was “entirely false.” But it also “almost started a war,” the source said. Any US operation against a Chinese vessel could have risked spiraling into an armed conflict between the two nations.

The Trump administration is considering proposals for at least 12 data centers and data center-related infrastructure projects to be built on federal public lands — far more than previously known.

AI is the road to mad money in the US, according to the business press. How mad is this money? Let’s turn to the Forbes 400 list of the wealthiest people in America.

The bar to join this list has never been higher, at $4.4 billion now. That’s an increase of $600 million, 60% of $1 billion, versus the past year. Living large has never been so, well, large.

Nice work if you can get it, no? In a capitalist society, the accumulation of such mind-boggling wealth is the highest priority for the ruling class. In 2026, the AI mad money grab is on.

AI is no benign thing that enables billionaires’ wealth. In fact, AI has grabbed much attention since an Anthropic researcher, Jacob Coxon, warned that AI “could kill us all by the end of the decade.” However, AI lethality is here now.

Take the Israeli military, with the financial and political backing of Uncle Sam, using AI to maim and murder Palestinian civilians in Gaza. For these victims, Israel’s use of AI is rampant, according to author Chris de Ploeg in his new book, The Exterminating Empire: Capitalism’s War on Life from Palestine to the Burning Planet (Monthly Review Press, 2026).

Further, de Ploeg writes: “Gaza has been turned into the world’s largest death camp ever to have existed. The murders are carried out, industrially, with bombs and algorithms.”

The Trump administration is considering construction of at least a dozen artificial intelligence data centers on thousands of acres of public land across six Western states, reporting revealed Friday, while a separate analysis found that Big Tech is receiving tens of billions of dollars in tax breaks for AI investments they were already likely to make.

The US Bureau of Land Management (BLM) is reviewing proposals for at least 12 data centers and related infrastructure projects on 17,600 acres of public land across Arizona, Idaho, Nevada, Oregon, Utah, and Wyoming, according to an investigation published Friday by The Washington Sun’s Mara Hoplamazian and Jade Lozada. Many of the proposals had not previously been publicly reported.

Classical data centers in space defy thermodynamics. Quantum AI in orbit compounds the impossibility. A technical teardown.

I have wanted to write about data centers in space since a startup called Lumen Orbit raised its seed round on the strength of a white paper and a rendering. I thought the idea was nonsense then, and I said so privately. I did not say so publicly, because orbital thermodynamics is not my field and I did not want to be one more voice piling onto a startup for the sport of it. That restraint lasted until July, when an investor I sometimes help with technical due diligence called and asked me about quantum AI data centers in space. The quantum half is my field. So here we are.

He wanted to talk through a hypothetical, or what he called a hypothetical, though I suspected he was evaluating an active pitch. The scenario: a startup proposing to build quantum AI data centers in orbital space. Superconducting quantum computers and GPU clusters on satellites, powered by solar panels, cooled by the vacuum of space, selling quantum computing and quantum AI as a service, including quantum-enhanced model training, quantum optimization, and quantum simulation workloads. He walked me through the concept in broad strokes, without naming a company or sharing a deck, and asked whether the physics worked.

It does not. I told him why, and he told me to write about it. He had given me no details, so there was nothing to keep confidential. This is also, inevitably, the article I have been not writing for two years, because before you can explain why quantum AI data centers in space are nonsense, you have to explain why classical data centers in space are, at best, a multi-decade engineering aspiration that the current wave of venture capital is treating as a near-term business plan.

Three layers of impossibility stack here.

Voters are anxious about losing their jobs to artificial intelligence, and key players across the political spectrum have started to notice. 

Now, the Working Families Party has rolled out a slate of policy proposals for the midterms, backed by more than two dozen Democratic candidates and representatives, that aims to address that anxiety. Their plan to counter AI-related job losses? Not a direct cash dividend, but a program seeking to place Americans in union jobs. 

A recent Quinnipiac poll showed that over half of Americans believe AI does more harm than good in their day-to-day lives, and 70 percent think that broad AI adoption will decrease the overall number of available jobs. 

With the midterms coming up, corporations and politicians are looking to address these fears. This month, OpenAI proposed creating a “public wealth fund” that would “provide every citizen with a stake in AI-driven economic growth.”

Yesterday, New York Assemblyman Alex Bores proposed a taxation framework designed to redistribute wealth from major AI corporations to people whose jobs might be displaced by their products, calling it an “AI dividend.” 

The Working Families Party, meanwhile, is proposing what looks like another Green New Deal-style jobs program to solve the same problem. 

Paramount Skydance is in "advanced talks" with California Attorney General Rob Bonta to settle the 12-state antitrust suit that's holding up its takeover of Warner Bros. Discovery, the WSJ reported.

The two sides have discussed Paramount potentially agreeing to run Warner Bros. studio separately for some period of time, per the Wall Street Journal.

However, Bonta has previously said that running the two companies separately — but under common ownership — would be a mere “behavioral” remedy, and that he is looking for a structural remedy that would entail separate ownership.

Scaleout deploys decentralized AI-driven learning to military bases and drones.

As European militaries adapt to the use of AI and drones in modern warfare, a NATO-backed startup is helping to deploy AI-driven target detection and selection that can run on small drones for surveillance and attack missions.

The company, Scaleout Systems, was originally founded by researchers from Uppsala University in Sweden in 2018, and initially focused on training and deploying machine learning models directly on the hardware available in commercial trucks and other vehicles. But once Russia launched its full-scale invasion of Ukraine in 2022, the company pivoted toward defense applications.

“With the war in Ukraine and a shifting world, we realized that this technology can be very important to operationalize edge data and sensor data for machine learning to make sure that NATO allies have found that strategic advantage,” Andreas Hellander, cofounder and CEO of Scaleout Systems, told Ars.

Instead of using frontier AI models from OpenAI or Anthropic, Scaleout is harnessing leaner ones, such as machine learning models that can perform computer vision tasks on the hardware of drones or computers used at forward bases. “They need to fit on forward-deployed hardware and edge hardware, which can vary quite a bit from small embedded devices to quite powerful edge workstations,” Hellander said.

OpenAI isn’t open.

The trillion-dollar artificial-intelligence firm has hidden its plans for local data centers behind a seemingly impenetrable wall of nondisclosure agreements and secrecy.

And it’s not alone. Other leading AI companies — including Amazon, Anthropic, Google, Meta and xAI — routinely require local and state officials to sign strict NDAs before discussing data-center impacts to local infrastructure and natural resources. The tech giants’ use of NDAs has become the latest flashpoint over data-center development in communities across the United States.

Often negotiated between government officials and the shell companies that build data centers, the NDAs are crafted to conceal the identity of the tech giants funding these sprawling energy-and-water-sucking complexes — allowing billionaire-owned tech firms to operate in the shadows.

For a time, many in the media portrayed data centers as a boon to local communities, bringing jobs and investment into regions that have fallen on hard times. But over the past year, that tale has turned. That’s due in part to the work of local independent news outlets that have engaged with local activists, organizers, community members and experts to expose the dangerous and often-illegal practices driving data-center development.

In Wisconsin, Wisconsin Watch has led much of this vital work. The outlet is a scrappy and independent nonprofit-news service dedicated to strengthening communities through engaged local journalism.