Tag: technology

Tech Giant Feared Its AI Tools Started A Commercial ‘Doom Loop,’ New Documents Reveal

AI executives allegedly knew their products could unlawfully harm businesses, flood the market with inferior work, and damage human society’s creative capacity as a whole, according to communications newly revealed in a lawsuit.

Those internal documents were cited in a lawsuit brought by The New York Times, Mother Jones, and other news outlets alleging OpenAI and Microsoft’s AI tools stole their work, and are harming their bottom lines.

“Millions of people around the world will soon consider large models ‘hoovering up’ all their work to be an astonishing theft of unprecedented proportions,” one Microsoft document said, according to a legal filing first reported by Mother Jones on Sept. 22.

The ‘Doom Loop’

One Microsoft document also allegedly indicated that the company was aware that the AI model outputs would result in fewer clicks on websites – including its own. Fewer clicks means less money, which disincentivizes creativity and leads to less, and maybe lower-quality, work.

“Our AI content strategy has started a ‘doom loop’ that will hurt the performance of our models and the entire web at the same time,” the document said.

In evaluating the suits, courts will look at how the new material produced by AI models affects the source material’s moneymaking capabilities.

According to the filing, Nick Turley, OpenAI’s head of ChatGPT, wrote that the company’s AI products pose an “existential threat” to publishers, “are largely substitutive, period” and “will get more and more substitutive as they get better.”

Those statements kill the AI companies’ “fair use” claim, attorneys for the news publishers wrote in their filing.

“The AI-generated outputs provide substitutive answers that reduce the need for users to visit Plaintiffs’ websites and flood the market with low-quality material that undermines the business model of news publishers,” the complaint reads.

One employee told OpenAI cofounder Greg Brockman that he’d developed “a hack” to get around The New York Times’ paywall, allowing ChatGPT to scrape up more stories.

“Ah, nice,” Brockman allegedly responded.

After the companies filed suit, the lawsuit says, OpenAI also developed a filter to stop ChatGPT from outputting from sources involved in the litigation.

“OpenAI did not suppress the output of content from any entity that had not sued it,” the lawsuit alleges. The filter was created “not to prevent OpenAI’s models from infringing copyrights, but to stop Plaintiffs from gathering evidence of OpenAI’s copying for use in litigation.”

An executive at Microsoft, the lawsuit claims, worried that might be an “accidental cover up” that would result in “people who have a right over the content having less visibility into what was used for training.”

Keep reading

Regulators Find Their Situational Awareness: Fed, BoE Probe Bank Exposure To Jane Street After AI Fund Meltdown

Better late than never…

Nearly two months after Leopold Aschenbrenner’s Situational Awareness went from a $45 billion AI juggernaut to a Citadel block trade after it dumped its entire public book on Ken Griffen’s doorstep at a 10% discount and six weeks after we learned that Jane Street lost $15 billion in July, its first down month in a decade, central bankers on both sides of the Atlantic have noticed that something may have happened.

According to the FT, the Bank of England’s Prudential Regulation Authority and the Federal Reserve have “stepped up scrutiny” of banks’ exposure to large trading firms and market makers. They are asking global lenders how much they have lent to Jane Street and Ken Griffin’s Citadel Securities, and presumably also to such HFT money makers as Susquehanna and Hudson River Trading.

Regulators want to know three things: the firms’ risk appetite, how banks’ exposure to them “evolved during the day,” and how risk controls held up. That’s a polite way of asking whether anyone at the prime brokers was actually watching intraday margin as the AI trade fell apart in July.

While the PRA, the Fed and Jane Street all declined to comment, Jane’s silence is the least surprising. Its last public word on the subject came from partner Turner Batty, who told investors, in the understatement of the year, that “July was a bad month.”

A brief history of a very bad month

For anyone who missed it (i.e., all the regulators), here’s the recap.

Situational Awareness is the fund Aschenbrenner, a former OpenAI researcher, launched in 2024 under the name of his viral essay. It had eight employees and ran long AI infrastructure / short software, with about 4x leverage through Goldman total return swaps. We described that leverage as “batshit insane”, but it was also extremely profitable, if only to Goldman. In a separate piece, the FT today reported that Goldman earned more than $200 million in fees this year from lending to Situational Awareness, the most of any client in its prime brokerage business financing hedge funds. 

As long as the market was going, up it was a party: the extremely levered momentum-chasing fund was was up 439% net through June, reached $45 billion in AUM, and counted Jane Street among its investors.

Then July happened. AI stocks rolled over, software rallied, and both legs of the pair trade lost money at once. Nebius, Sandisk and SharonAI each fell roughly half, and SK Hynix dropped nearly 50%. As the margin calls came in, the fund went looking for fresh capital. It failed to find willing “widows and orphans” and so within days it had exited all of its public equity trades. Citadel bought the whole book in under 24 hours at about a 10% discount, beating Millennium and, fittingly, Jane Street. The fund ended the month down about 78%, with roughly $10 billion of private holdings left, including a large stake in Anthropic.

Griffin did well out of it. Citadel’s Wellington fund rose 5.94% in July, its best month since 2022, and about half of its year-to-date gains came from the Situational Awareness trade. Over the next three weeks, Citadel flipped more than 80% of Leopold’s portfolio to dumb money through roughly 100 block trades worth more than $4 billion. When Griffin later described the unwind, he thanked “the trading and prime brokerage teams at the banks serving both firms” for their “extraordinary cooperation.”

Regulators are now asking those same prime brokerage teams some questions of their own.

Jane Street was the collateral damage: on top of its direct stake in Situational Awareness, its own book, which tends to, cough, be just ahead of whale and retail orders, leaned into the same momentum names: Sandisk, Micron, CoreWeave, Broadcom, SMCI, Dell and Bloom Energy, all the names that defined the momentum trade trough July. And when the AI trade reversed, it lost $15 billion in a month. Most firms wouldn’t survive a hit like that. At Jane Street it barely dented the year: by early August the firm had generated $40 billion in net trading revenue, already more than its record $39.6 billion for all of 2025.

That is also what worries regulators. As the FT puts it, the size of the loss “indicated that Jane Street… appeared to take far more risk than a typical market maker.”

Keep reading

Flock And Other Tech Firms Decline Senate Testimony On Mass Surveillance

Flock Safety CEO Garrett Langley and the heads of Axon Enterprise, Motorola Solutions and Verkada all declined invitations to testify before a Senate Judiciary subcommittee on Wednesday, as senators from both parties pressed on the surveillance networks their cameras feed. Sen. Josh Hawley (R-Mo.), who chairs the Subcommittee on Crime and Counterterrorism, called the no-shows “so typical.”

Democrats weren’t any happier. “For the Flock CEOs, obviously they’re nervous about being on camera,” said Sen. Dick Durbin (D-Ill.). Sen. Richard Blumenthal (D-Conn.) went further: “Their refusal to come forward and answer questions is a sign that they have something to hide, and they are refusing to be accountable.”

Instead, the panel heard from Lindsey Isaacs, a Florida woman who spent 13 days in jail after Flock’s system wrongly flagged her car as involved in a fatal crash. She was charged with three counts of vehicular homicide. The charges have since been dropped.

As The Epoch Times notes further, While the license plate-reading cameras developed by Flock have gotten the most airtime in recent weeks – prompting nationwide, viral moments of activist vigilantes destroying or disabling the cameras amid public opposition to the technology – competitors at Axon Enterprise and Motorola offer similar options for police and local governments.

The Los Angeles Police Department is currently planning to roll out a 10-year contract with Axon, a $235 million arrangement that would equip America’s second-largest metropolis with more than 1,600 license plate readers.

For municipalities, companies like Flock, Axon, and Motorola offer a potential melding of technology and traditional surveillance tech to reduce crime and make roadways safer. But the cameras they operate have boosted a growing artificial intelligence-powered surveillance network accessible by thousands of U.S. law enforcement agencies, which critics say runs afoul of the Fourth Amendment’s protections against unreasonable search and seizure.

Hawley warned during the hearing that, if not addressed, this would have troubling implications for Americans.

“This is still the United States of America, and it shouldn’t be that when you walk out your door in the morning, you just have to agree in assent and accept the fact that you’re going to be on camera morning, noon, and night,” Hawley said.

“You’ll be tracked by anybody and everybody in the world. You’ll have no privacy whatsoever. You’ll be tracked when you take your kids to school and tracked when you go to church and tracked when you go to the grocery store.”

Some high-profile abuses of the data collected from these cameras have made headlines.

In one recent case, Indianapolis authorities announced that five police officers had been charged with misusing data from Flock.

Keep reading

Proper Flock’d: What This Cop Did Raises Major Red Flags…

A Naperville police officer turned a statewide camera grid into a personal tracking app, ran a man’s plate about 150 times, and then retired like the file never existed.

The story serves as yet another example of how the burgeoning mass surveillance grid is being abused on a regular basis.

Patrick Dowhen was not a suspect. He was the ex. After Officer Oscar Maldonado started a relationship with Dowhen’s fiancée, Flock became the shortcut. No warrant. No case number that mattered. Just a badge, a login, and a national lattice of plate readers that can reconstruct where you went, when you went, and how often you go there.

“Anywhere I went in the country, I was being tracked,” Dowhen told ABC7 Chicago’s I-Team. “It was scary, it was really scary.”

Internal records obtained by the I-Team show the pattern started after Maldonado’s relationship with Dowhen’s then-fiancée began. Investigators found evidence he searched Flock for Dowhen’s vehicle approximately 150 times over a three-month stretch last year.

Chicago Tribune reporting, from a separate FOIA haul, put the plate runs between June 18 and Sept. 15, 2025, and said Maldonado tracked plates for non-police reasons hundreds of times.

Dowhen said the surveillance was not abstract. “My ex-fiancée would tell me where I was at, or what I was doing, and he [Maldonado] was feeding her information,” he said. “He’s an officer, feeding a civilian, you know, inside information.” He tore his truck apart looking for a tracker. Then he went to Naperville Police Chief Jason Arres. That complaint opened the internal case last September.

Maldonado’s story to interrogators was protection: the woman, himself, his family. Naperville investigators did not buy it. They wrote that the “evidence contradicts Maldonado’s explanation that he only used Flock for the safety of himself, his family,” and Dowhen’s ex-fiancée.

The report also concluded there was “a preponderance of the evidence that shows Maldonado did commit the crime of Official Misconduct,” adding that he “admitted to using FLOCK for personal reasons.”

Then on May 1, 2026, Maldonado retired. Department spokesman Rick Krakow said the internal case was complete but not closed because discipline had not been imposed. “However, the officer resigned before his disciplinary meeting with the chief, which essentially closed the case.”

No arrest. No plate charges. A referral to the Illinois Law Enforcement Training and Standards Board. Will County prosecutors would not confirm or deny a criminal review. Dowhen later obtained a stalking no-contact order that noted Maldonado “was tracking my plate since 06/2025.” That order was dropped after both men agreed to a two-year mutual no-contact deal.

“He could very well get a job as a police officer again and have, you know, all these same electronic things available to him at any time and stalk anybody you want,” Dowhen said, “because at this point he has no criminal record.”

ABC7 chief legal analyst Gil Soffer laid out the gap that let a 150-search campaign end in a retirement party. “It’s not enough to say that a police officer violated internal laws, you’d have to show that a police officer knowingly violated a law, a statute, a constitutional provision,” he said.

He added, “An official misconduct statute in Illinois won’t necessarily cover the simple situation where a police officer accesses license plate information for an improper purpose.”

Maldonado’s own lawyer, Jeff Tomczak, made the same point from the other side of the table. “I’m a defense attorney. It should be a crime. It should absolutely be a crime. The legislature should act, create the crime of misuse of Flock. Don’t require the state’s attorneys to be creative.”

A private company sold cities a warrantless movement database. Departments wrote policy memos. The statute book never caught up. When an officer treated the grid like a jealous-ex app, the fallback was an employee manual and a quiet exit.

The I-Team says this is not one suburban rogue. It identified 14 Chicago-area cases of officers accused of misusing Flock, with at least five sent to prosecutors. Berwyn’s mayor, Robert Lovero, issued an executive order this week suspending the city’s cameras.

Keep reading

Report: U.S. Voters Turning to Chatbots for Election-Related Information

American voters are turning to artificial intelligence (AI) chatbots and AI-generated search summaries for election-related information in growing numbers, according to research published in 2026 by the Institute for Strategic Dialogue (ISD) and the AI firms Orchestra and Anthropic.

According to a report by The Hill, the trend spans both dedicated chatbots and AI overviews embedded in search engines. U.S. adult chatbot usage rose from about one-third of adults in 2024 to about 50% in 2026, per a study commissioned by Orchestra [1]. About 650,000 chatbot prompts reviewed by the firm related to politics, the study found, with ChatGPT and Google AI tools the most frequently used sources, according to the research [1].

The shift is drawing scrutiny from election officials and researchers because AI systems can generate answers that are incomplete, outdated or inaccurate, the reports stated. The findings arrive as the 2026 midterm elections approach, with state and local offices preparing for voters who may arrive at the polls with information assembled by automated tools [2].

Use of AI for Political Questions Grows

An analysis by the Anthropic Economic Index found that politics ranked No. 66 out of 189 topics, a level the researchers described as “modest, but still meaningful,” according to the report. In states holding May primaries, political conversations on the platform rose about 20%, according to Anthropic’s Andy Hall and UCLA’s Dan Thompson, who authored the analysis.

“If primaries can lift political AI use this much, November’s midterm elections will likely bring the biggest wave of political AI use on record,” Hall and Thompson wrote. The projection suggests that general-election traffic will exceed the primary-season increase, according to the report.

Max Read of the ISD said AI outputs “are going to be playing an increasing role” as a primary information source and in subtler ways, according to The Hill [2]. The Associated Press reported in a May investigation that AI “can not only rapidly produce targeted campaign emails, texts or videos; it also could be used to mislead voters, impersonate candidates and undermine elections on a scale and at a speed not yet seen,” according to the report [2]. The World Economic Forum’s Global Risks Report ranked misinformation and disinformation as the foremost threat to humanity in its 19th edition, ahead of climate change, the report stated.

Accuracy and Completeness Concerns Persist

An ISD report found that 29% of election-related chatbot replies were incomplete, inaccurate or outdated, according to the report. Errors included misidentifying Election Day dates and citing outdated requirements, the report stated. Another 16% of responses answered correctly but omitted useful details such as deadlines or ID requirements, according to the findings.

Accuracy dropped 16 percentage points when prompts were submitted in Spanish, the report stated. The language gap raises questions about whether AI tools serve voters with limited English proficiency as reliably as English-speaking users, according to the researchers.

Ben Worthen of Orchestra said the growth in people turning to these tools for information questions has been “staggering,” and that Google AI overviews now answer much of what users type into Google. Independent researchers have documented similar patterns across other domains, including public health, where chatbots have generated flawed medical advice according to independent research [3]. A study by the Digital Media Research Center found that AI chatbots often fail to debunk conspiracy theories, instead presenting them as plausible alternatives and engaging in “bothsidesing” by offering false narratives alongside facts without clear refutation [3].

Keep reading

Humanoid Robot Floors Influencer in History-Making MMA Fight

A humanoid robot sent an influencer sprawling in a history-making MMA bout, offering a striking demonstration of how rapidly machines are acquiring the strength, balance, and agility once thought to provide a measure of job security for human workers.

Content creator Frankie LaPenna, 29, faced the 6-foot-tall “Rekbot” in a Sept. 18 bout staged by robotics startup REK and billed as the first human-versus-robot MMA fight.

The widely seen video showed the two trading blows before the Terminator-like machine landed a powerful kick that knocked LaPenna across the ring.

The robot was remotely controlled by a human rather than acting autonomously.

Perhaps more remarkable than the punches and kicks was Rekbot’s ability to recover.

When knocked down or struck off balance, the humanoid could rapidly spring back to its feet, displaying the kind of dynamic balance, coordination, and physical resilience that until recently remained one of robotics’ biggest obstacles.

“We just made history with the first Human VS Terminator robot fight,” REK wrote after the event, saying LaPenna “gave it his all.” The company later teased that Rekbot “wasn’t even the final boss.”

According to a post on X by Rek, Lapenna “kept going and defeated” Rekbot.

The spectacle comes as the debate over automation is expanding beyond artificial intelligence’s potential impact on programmers, writers, accountants, and other white-collar professions.

Generative artificial intelligence itself still has major limitations when physical execution is required, but increasingly sophisticated robotics could give AI something it previously lacked: a capable body.

That transition is already moving from the laboratory to the factory floor. Boston Dynamics says its production-ready Atlas humanoid is designed specifically for industrial work and is being deployed in 2026.

This week, the company opened a facility at Hyundai’s Georgia auto plant where Atlas robots are being trained on real manufacturing tasks, including handling and sequencing automotive parts for assembly.

The implications could eventually extend to jobs once considered comparatively resistant to automation because they demand strength, mobility, endurance, and dexterity.

The World Economic Forum said in its 2025 jobs report that robots and autonomous systems were expected to be the largest net technological job displacer through 2030, with an estimated net reduction of 5 million jobs.

Employers surveyed also expected the importance of “manual dexterity, endurance and precision” to decline as robotics and other technologies automate more physical tasks.

That does not mean humanoids are presently capable of replacing tradespeople broadly, as current systems remain expensive and limited compared with humans operating in unpredictable environments.

But the Rekbot demonstration illustrates how quickly one long-standing technical barrier is eroding.

Machines are no longer merely lifting objects in repetitive motions from fixed positions on assembly lines.

They are increasingly capable of balancing, recovering from falls, and executing complicated whole-body movements as well as making instant decisions in new environments.

The fight was staged as entertainment to promote REK’s virtual-reality robot game.

But the impressive visuals of a machine taking a hit, hitting the floor, and immediately springing back up may prove more consequential than who won the bout, as it shows AI is advancing into cognitive work at the same time that increasingly agile robots are advancing into the physical world.

Keep reading

An AI Went Looking Through DNA and Found Something Humans Had Missed

For years, we’ve been asking whether artificial intelligence can write like us, draw like us, code like us, or eventually replace us. Anthropic just gave us a more interesting question: Can AI notice something in nature that human scientists have been staring past?

On Wednesday, Anthropic announced that its Claude AI had identified what researchers believe is a previously unrecognized biological system hidden in the DNA of bacteriophages, viruses that infect bacteria. The system contains a reverse transcriptase, a nearby partner gene, and a long array of repeating DNA sequences. Anthropic calls it ART, short for array-associated reverse transcriptases.

From Anthropic:

Today, we’re sharing early results from one of our first research programs, in which Claude autonomously discovered a novel enzyme system that is associated with an array of DNA repeats, a pattern reminiscent of CRISPR. Although we don’t yet know its function, the system that Claude discovered has a set of characteristics that have only ever been found together in a handful of other systems, all of which are programmable and perform operations like cutting, copying, and pasting DNA. Beyond CRISPR, which has already transformed science and medicine, several other such systems are now in development as promising tools.

The system that Claude found is based on a reverse transcriptase (RT), enzymes that copy RNA into DNA. While this underlying RT, found in a jumbo phage, had been identified in previous studies, Claude appears to be the first to notice the system’s defining features—an associated array of non-coding DNA sequences and an additional accessory protein of unknown function.

The enzyme itself wasn’t unknown. Scientists had already catalogued the reverse transcriptase. What humans apparently missed was the larger pattern surrounding it. Claude noticed that these pieces occurred together and that the repeating DNA looked strangely organized.

Keep reading

AI could harm children more than social media – ex-Google exec

Artificial intelligence could pose greater risks to children than social media, with tech companies failing to put adequate safeguards in place, a former Google safety executive has warned.

Tom Siegel, who founded Google’s trust and safety team and spent 15 years at the company, told Reuters this week that the industry was repeating mistakes made during the rise of social media.

“We find ourselves in a really terrible situation in terms of what AI is doing to kids,” Siegel said. “The idea that the industry will figure it out by itself without outside help is unrealistic.”

He warned of “very severe scenarios of harm,” citing risks ranging from suicide and psychosis to ‘cognitive offloading’ – reduced critical thinking caused by relying on AI for answers. Companies can and should take immediate steps, including stronger age checks to prevent children from accessing chatbots, he said.

Keep reading

Russia Guts Ukraine’s Drone Industry Overnight — And NATO Scrambles – What They Won’t Tell You

Russian precision strikes on Ukrainian drone manufacturing facilities were carried out in a massive overnight operation to destroy a significant part of Ukraine’s drone production capacity, while NATO fighter jets were scrambled in Poland and Romania.

The Russian Ministry of Defense reported that its armed forces delivered a massive strike using ground and air precision weapons against Ukraine’s military-industrial complex, telecommunications centers, logistics hubs in Kyiv and the Odessa region, and naval vessels supporting Ukrainian forces.

Among the confirmed targets: an industrial facility manufacturing long- and medium-range UAV components and ground robotic systems; the Pegasus Arms LLC workshop, which stored and produced fixed-wing drones; and the Element UA LLC facility producing Hornet strike drone components. A Kyiv transport and logistics center for distributing long-range UAVs was also struck, as were the Kievstar and Parkovy telecommunications centers providing communications and computing support to Ukrainian forces.

In Odessa, two military logistics centers and two Ukrainian naval vessels were hit.

Ukraine reported Russia launched 161 strikes overnight — 82 jet-powered drones and an unknown number of Banderol cruise missiles. Air defenses intercepted 119 drones and three cruise missiles, with impacts recorded across 18 locations.

Poland scrambled fighter jets after Russian strike drones operating in western Ukraine approached its airspace. The alert lasted approximately an hour; no violations were recorded. Romania scrambled two F-16s to respond to an aerial target maneuvering near its Ukrainian border — also with no incursions confirmed.

The NATO scrambles have become nearly a nightly occurrence. Russian drones have previously crossed into NATO member airspace and crashed on NATO territory.

Keep reading

Drones Are Bringing The Surveillance State To Your Front Door

At the moment, the United States government has no official public count of how many drones law enforcement agencies own and fly in the U.S. There is no list from the Federal Aviation Administration (FAA), the Department of Justice (DOJ), or any state agency accessible to the public. Various industry estimates put police drone programs in the thousands — one recently cited as high as about 6,000. Congress should not have to take its numbers from corporate executives on quarterly earnings calls about how many drones are already circling overhead and surveilling the American people. The Department of Transportation, FAA, DOJ, and the Office of Justice Programs should put the fleets on the books and brief the House and Senate.

Until recently, most drones sold to police departments were made in China, raising significant concerns about Chinese government access and influence — either directly from China or through American companies that marked them up and profited. Congress has spent years trying to stop the Chinese drone invasion of our police departments. Those efforts were necessary. They are not finished. Even if new sales slow, fleets of Chinese drones remain deployed across the nation. And there are still continued efforts to keep Chinese components, software, and supply chains in the fleet.

Not one piece of hardware or software from the People’s Republic of China belongs on a police or federal drone in the United States, or in any piece of related equipment for that matter.

Not the chip, camera, battery, flight controller, app, firmware, or cloud login.

Grant funding, FAA registration, and operational approval should not be granted to aircraft or ground systems that contain Chinese hardware or software. American first responders deserve American supply chains.

That is already the direction of federal law, and it ought to be enforced without loopholes. Section 848 of the National Defense Authorization Act for Fiscal Year 2020 barred the Department of Defense from operating or contracting for covered foreign unmanned aircraft systems and critical components. The American Security Drone Act of 2023 was enacted as part of the National Defense Authorization Act for Fiscal Year 2024. The order on funding Unmanned Aircraft Systems says no Office of Justice Programs funds may go to unmanned aircraft systems manufactured or assembled by a covered foreign entity, with only limited exceptions.

Members of Congress should not have to guess how many drones are flying overhead. Congress must investigate and coordinate with executive agencies to determine the facts: how many unmanned aircraft federal agencies and state, local, tribal, and territorial law enforcement agencies already fly; fleet size by agency; their missions and payloads; and under what rules, if any, they operate. That is the level of transparency we need for government-used drones. Existing fleets should be registered within 30 days of operational deployment so we can put the brakes on unfettered overhead surveillance.

Federal tax dollars should go to real emergencies — search and rescue, wrecks, officer safety, and a barricaded suspect — not ever-present aerial patrol above American homes.

Those dollars already flow through the Office of Justice Programs, the Community Oriented Policing Services Office, Byrne Justice Assistance Grants, among others. Unmanned aerial systems are an explicit priority for Fiscal Year 2026 Community Oriented Policing Services Community Policing Development. “Drone as first responder” is not a substitute for a warrant. Persistent neighborhood surveillance is a general warrant from the sky, and a microphone on a drone is an intercept. Listening should require the same legal process as a Title III wiretap under 18 U.S.C. §§ 2510–2523, or the state equivalent — no fishing expeditions, audit logs, and short retention.

Weaponized drones bought with Department of Justice grant money are already banned. That ban should apply to all related funding streams unless Congress authorizes otherwise.

Keep reading