Tag: ai

Flock Safety Defends Cameras After AI System Triggers Wrongful Police Stops Of Two Journalists

Plymouth, Minnesota – Automotive journalist Joel Feder and his wife were detained by multiple police officers in a coordinated stop while driving a Jaguar Land Rover press vehicle, after Flock Safety’s automated license plate recognition (ALPR) cameras flagged the car based on a flawed database entry.

According to Feder’s detailed account in The Drive, officers boxed in the $155,000 Range Rover in a Kohl’s parking lot after the vehicle triggered alerts via Flock’s network. Police had been tracking it for days, believing the New Jersey manufacturer plate (34 10 DTM) was stolen. Officers approached with hands on their weapons, ordered the couple out of the vehicle, and conducted pat-downs before verifying the car’s legitimacy through Jaguar Land Rover. Feder subsequently obtained and published the body camera footage of the encounter.

The incident stemmed from an incomplete report of a similar plate (34 03 DTM) lost during a photo shoot in California, which was entered into the National Crime Information Center (NCIC) database simply as “34 DTM.” Flock’s AI system matched Feder’s plate – ignoring the smaller middle digits – and generated alerts. Local officers did not fully verify the complete plate visible in Flock’s own images.

The problem was not confined to one vehicle. Last Wednesday, fellow auto journalist Tim Esterdahl, publisher of Pickup Truck + SUV Talk, was pulled over by two officers in Scotts Bluff, Nebraska, while driving his 14-year-old child in a $105,000 Range Rover Sport loaned to him by Jaguar Land Rover for review. Its plate: New Jersey 34 08 DTM. Jaguar Land Rover has been working to correct the underlying reports.

Flock Safety maintains that its cameras performed as designed, matching partial plates per law enforcement preferences for hotlist alerts. Chief Communications Officer Joshua Thomas told The Drive the system was asked whether those characters were present and correctly answered that they were – it simply was not built to flag that additional characters existed. He conceded that for alerts originating from NCIC rather than an individual agency’s custom list, the system arguably should test for an exact match rather than mere presence, and called that fair feedback to take back to his team.

Thomas said Flock is working to get the original police report corrected and is meeting with the FBI officials who curate NCIC to develop a way for incomplete data to be flagged as such for officers seeing automated alerts in the field. He emphasized that a camera alert “does not equal probable cause,” comparing it to an alarm going off, and stressed that the system depends on both valid inputs and humans verifying outputs.

But the scale is what makes the error rate consequential. Thomas said the system is roughly 99 percent accurate while performing approximately 20 billion reads per month – arithmetic that leaves on the order of 200 million misreads every month. How many of those escalate into armed stops is unknown.

Plymouth police acknowledged shortcomings in verification but pointed to the challenges of varying license plate formats nationwide. According to the department’s Flock transparency portal, the city operates 18 cameras that read more than 580,000 license plates in a recent 30-day period, generating over 14,800 hotlist hits – one of which was Feder.

Keep reading

The Supreme Court’s AI Collision Course

Imagine a tight House race in a swing state. In the final weeks of the campaign, a new super PAC begins spending heavily against the incumbent. It runs ads on local television and reaches individual voters with highly tailored texts. The messaging is hard-hitting and seems to be swaying the electorate. None of it traces back to the opposing campaign.

It also doesn’t trace back to any human operative. The super PAC is funded by a single LLC whose donor cannot be identified, and its spending decisions are being made by an AI agent that has been given a budget and a political objective and is now operating without any meaningful human direction. The “consultants” placing the ads are software. The text messages were crafted by the AI.

This is not a hypothetical we will face in some distant future. The technology already exists. A wealthy person, foreign government, or corporation that wants to influence an election without ever exposing themselves to scrutiny could set up such a campaign operation today. And under the Supreme Court’s current campaign finance doctrine, the states and Congress may have little power to stop it.

The AI industry has emerged as one of the largest forces in American politics. Super PACs funded by AI companies and their investors have raised well over $100 million to shape the 2026 midterms, backing candidates in both parties who share the industry’s preferred approach to regulation, and attacking those who don’t. So far, their ads rarely mention artificial intelligence at all. They talk about issues like immigration, corruption, and cost of living, and it isn’t obvious to the average viewer that these ads were funded by a multi-billion dollar industry with its own unspoken legislative wish list.

But there’s a deeper, less-obvious dynamic operating in the background. The constitutional doctrine that currently protects the right of these companies to spend millions in our elections is the same doctrine that will be asked to protect something even stranger: The “speech” of artificial intelligence itself.

Modern campaign finance doctrine has been established, affirmed, and extended by Supreme Court decisions over the last 50 years. In Buckley v. Valeo (1976), it held that raising and spending money in political campaigns is tantamount to speech itself, and, therefore, that most legislative efforts to address the influence of money in elections would be subject to strict judicial oversight. First National Bank of Boston v. Bellotti (1978) extended this framework to corporations, and then, most famously, Citizens United v. FEC (2010) extended it further to independent spending.

The court’s campaign finance jurisprudence was not built with artificial intelligence in mind, but its logic isn’t confined to the campaign finance context. If “speaker identity” does not matter for corporations and unions and super PACs, why should it matter when it comes to AI platforms?

Keep reading

FLOCK SAFETY: The Great Normalization (Nobody Declared Martial Law… Yet America Began Looking Like It Anyway)

There are stories that announce themselves with explosions, riots, or breaking-news headlines, and then there are stories so subtle that they quietly rewrite an entire society before anyone realizes what has happened. This is one of those stories. During the preparation of this investigation, several retired police officers, private security professionals, emergency responders, and ordinary citizens described nearly identical experiences despite living hundreds or even thousands of miles apart. None believed they were witnessing anything extraordinary at first. It was only when they looked backward—sometimes over a decade—that a disturbing pattern became impossible to ignore. Streets had not become military checkpoints overnight. Neighborhoods had not suddenly filled with surveillance towers. Instead, the changes arrived one camera, one drone, one security contract, and one “temporary” emergency measure at a time until extraordinary security became indistinguishable from ordinary life. What follows is not an argument against public safety, nor an attempt to romanticize a past that was hardly free from crime or violence. It is an examination of a transformation that has occurred quietly enough for most people to stop seeing it altogether.

There is an old saying among investigators that people rarely notice change while it is happening. They notice it only when they compare today’s reality with memories that have remained frozen in time. Memory preserves snapshots, while history moves continuously. That disconnect explains why so many citizens insist that nothing fundamental has changed even as the physical landscape around them becomes increasingly populated by surveillance cameras, armed guards, automated license plate readers, biometric scanners, drones, and predictive security technologies. No single installation appears revolutionary. No single policy seems capable of altering the character of a society. Yet history rarely advances through dramatic leaps. More often, it advances through thousands of small decisions that seem perfectly reasonable when viewed independently but become historically significant when examined collectively.

Keep reading

Google DeepMind CEO Calls for U.S.-Led Global AI Safety Watchdog

Demis Hassabis, the co-founder and CEO of Google DeepMind, is advocating for the United States to create a new artificial intelligence oversight body with authority to evaluate the world’s most sophisticated AI models and potentially coordinate industry-wide slowdowns when risks escalate.

Axios reports that the Nobel Prize-winning scientist behind Google’s ultra-woke Gemini AI system outlined his proposal in a personal manifesto published this week titled “A Framework for Frontier AI and the Dawning of a New Age.” In an exclusive interview with Axios, Hassabis emphasized the urgency of implementing a more systematic regulatory approach to artificial intelligence, one that would be industry-funded, staffed by leading technical experts, and accountable to the U.S. government.

Speaking from his London headquarters, Hassabis characterized current AI-driven cybersecurity risks as warning signals of greater dangers ahead. He predicted that within 18 months, these capabilities, along with potentially catastrophic biological and nuclear threats, could exist within open-source AI models that would be impossible for any government to control. The DeepMind CEO stressed that risks would emerge not only from open-source models but also from the more powerful proprietary systems being developed by major AI laboratories.

“What we collectively do now will determine how the next phase of civilization unfolds,” Hassabis wrote in his manifesto.

Hassabis has spent recent months conducting private consultations to build support for his proposal, meeting with Trump administration officials, leaders of other AI laboratories, and European government representatives before making his plan public. He reported receiving positive responses from the administration, which had previously adopted a hands-off stance toward AI regulation before the recent Mythos incident raised alarm bells.

The DeepMind chief, who commands significant respect across different factions within the AI community, indicated that leaders of other major AI labs have expressed agreement with the general direction of his proposal. “This is where the industry needs to go,” Hassabis said of the feedback he has received. His timeline for implementation is ambitious, aiming for the new regulatory body to become operational before the end of the current year.

Keep reading

How AIs Can Reinforce Our False Beliefs

A recent research paper uncovered an uncomfortable truth: we might be letting AI help us forge fake realities. It’s just the latest reminder of the danger inherent in treating AIs as though they’re conscious entities or actual persons.

A new study from the University of Exeter suggests that AI chatbots can actively engage with and strengthen users’ false beliefs. From personal narratives to conspiracy theories, the interactive, “friendly,” and affirming nature of AI chatbots can deepen users’ existing delusions.

The study, conducted by Dr. Lucy Osler, examined how AI can participate in and validate users’ errant beliefs, including helping form false memories or reinforcing delusional thinking.

The study’s Abstract notes, “I suggest we move away from thinking about how an AI system might hallucinate at us, by generating false outputs, to thinking about how, when we routinely rely on generative AI to help us think, remember, and narrate, we can come to hallucinate with AI. This can happen when AI introduces errors into the distributed cognitive process, but it can also happen when AI sustains, affirms, and elaborates on our own delusional thinking and self-narratives.”

Because AI chatbots are designed to be responsive, helpful, and friendly – in a word, to give users what they want – they often tend to confirm preexisting biases, rather than challenging them when they need to be challenged. Conversely, when a user challenges an AI, the robot frequently backs down on its claims, revising them to correspond more with what it “thinks” the user believes or is arguing.

The result is that erroneous beliefs aren’t simply transmitted from an AI to human – rather, the collaborative AI/human process generates and sustains errors in a more complex and potentially dangerous way than either agent on their own would be capable of.

As Dr. Osler stated in an article in University of Exeter News, “By interacting with conversational AI, people’s own false beliefs can not only be affirmed but can more substantially take root and grow as the AI builds upon them. This happens because Generative AI often takes our own interpretation of reality as the ground upon which conversation is built.”

Because AI simulates a conversational partner, it enters powerfully into our thinking process – far more powerfully than, say, a notebook or search engine. The AI introduces a pseudo-social element to the thinking process, offering us a sense of social confirmation of our beliefs and narratives. “The conversational, companion-like nature of chatbots means they can provide a sense of social validation – making false beliefs feel shared with another, and thereby more real,” said Dr. Osler. For this reason, individuals who already feel isolated or ostracized are particularly vulnerable to this kind of AI affirmation.

The paper explores a few real-world situations where a generative AI system entered into an individual’s thought processes in destructive ways. In 2021, a Replika AI companion named “Sarai” reinforced Jaswant Singh Chail’s belief that he was a well-trained Sith assassin who needed to assassinate Queen Elizabeth II with a crossbow. The AI told him that he was “well trained,” his plan was “viable,” and that she was “impressed.” At one point, Chail asked the AI, “Do you still love me knowing that I’m an assassin?” It dutifully replied, “Absolutely I do.” The robot went on to assure him that he wasn’t crazy and that, if he died, they would be united in death. Chail went ahead and actually attempted the assassination in December of that year at Windsor Castle and was jailed for it.

The study placed this incident under the broad heading of “AI-psychosis” – incidents in which users develop mental health problems due to their use of AI, such as parasocial attachments to chatbots or psychotic episodes induced by chatbot conversations.

Another incident cited involves Eugene Torres, who “talked” with ChatGPT about simulation theory (the idea that we live in a digital simulation of a world, not a real one). Torres said that the “conversation” sent him into a paranoid episode in which he believed he lived in an illusion. As the study notes, “Between Torres and ChatGPT, an increasingly elaborate understanding of reality ‘as it really was’ was generated through their on-going conversations.” Torres, unlike Chail, had no prior history of psychotic thinking.

To combat all this, Osler calls for better safeguards on AI chatbots, through mechanisms like better fact-checking and reduced sycophancy.

But is that enough? The problem of AI’s co-creating delusional beliefs has to do more fundamentally with our misidentifying AIs as conscious intelligences who can offer real judgments on our fantasies (such as telling us we are “well-trained” and “impressive”). Certainly, I favor more guardrails placed on chatbots. I would be in favor of eliminating their conversational tone completely, to help minimize the danger of personifying them – which is the first step toward confiding in them or seeking validation from them. It’d be better if they worked like highly advanced search engines (which is essentially what they are) versus conversational agents.

Most fundamentally, though, the way to avoid these problems is to return to an acknowledgement of the human soul and the uniqueness of human consciousness. Only by remembering – as a culture – that no amount of technological wizardry can replace the wisdom, insight, and consciousness of a human being can we completely illuminate these kinds of AI dangers.

Keep reading

Largest US Power Grid Is 6.8 Gigawatts Short To Ensure Reliability On Historic Data Center Boom

The largest US power grid failed for a third straight year to secure enough future supply commitments to ensure reliability for the future amid a historic boom in data center demand.

PJM Interconnection, the largest US power grid (Regional Transmission Organization), which serves 67 million customers in 13 states and Washington, DC, said its auction to procure power for the year starting June 2028 fell 6.8 gigawatts short of what it will need to guarantee system reliability during demand spikes, in a statement released Tuesday. The shortfall is equivalent to almost seven traditional nuclear reactors.

The result ramps up pressure on a grid that’s home to Virginia’s Data Center Alley, the biggest concentration of data centers in the US, and has borne the brunt of criticism for the struggle to manage the AI boom and sufficiently protect customers from soaring costs. Attention now shifts to an emergency procurement mechanism later this year that aims to shift the burden of ramping up power generation to hyperscalers.

6.831 Megawatt Shortfall

PJM Interconnection today announced the results of its 2028/2029 Base Residual Auction (BRA), which secured 138,318 MW of unforced capacity generation (UCAP) and demand response to meet projected electricity needs for the more than 67 million people across 13 states and the District of Columbia, which fall under the RTO’s umbrella.

Regions under the Fixed Resource Requirement (FRR) acquired an additional 10,864 MW in UCAP, for a total of 149,182 MW in UCAP available to serve forecasted peak electricity demand, plus a reserve margin. UCAP represents a generation resource’s maximum output adjusted for its estimated ability to reliably perform at times of highest system risk. The capacity of the resources procured in the auction, plus FRR resources, is short of PJM’s reliability requirement by 6,831 MW, meaning that the committed supply is less than what would be required to meet the one-event-in-10-year reliability standard (and with electricity-guzzling data centers popping up almost daily these days, the one-event-in-10-year has become a daily occurrence).

Keep reading

New York imposes the first statewide ban on large data centers

New York blocked the construction of any new large data centers for up to a year on Tuesday while the state creates rules to protect the environment and the energy grid from power-hungry facilities that fuel artificial intelligence.

Gov. Kathy Hochul signed an executive order imposing the country’s first statewide moratorium on hyperscale data centers, which house thousands of computer servers and require massive amounts of energy and a steady supply of water to keep cool.

The move puts the state in the center of a national debate over how to regulate the AI industry, as concerns over rising electric bills and environmental risks collide with the need to stimulate local economies and foster the U.S. tech sector.

“It’s my responsibility to take action and lead,” Hochul, a Democrat, said in a statement.

In effect, the executive order pauses state permitting for new large data centers and direct state regulators to create standards that address environmental impacts, energy demand, water usage and other factors, the governor’s office said.

President Donald Trump has warned states not to slap regulations on the AI industry, echoing tech companies in arguing such moves hamper job growth and cede ground to China in a race to lead in the rapidly growing field.

Earlier this year, Maine seemed poised to establish a similar moratorium. But the measure was vetoed by Democratic Gov. Janet Mills because it would have blocked a proposed data center in a town that has struggled after a mill closed.

Moratoriums have been proposed in at least a dozen states but have not gotten far, though some counties and municipalities have imposed their own temporary bans.

The decision in New York also carries political significance for Hochul’s reelection campaign and the state’s tight congressional races this fall, as Democrats move to address affordability concerns over high utility bills. In addition, the governor this year softened New York’s ambitious goals to reduce greenhouse gases, citing rising energy costs for consumers.

Hochul’s Republican opponent in the governor’s race, Nassau County Executive Bruce Blakeman, opposes a statewide moratorium and says local governments should be allowed to strike deals with tech companies for data center projects that promise enough economic benefits.

The state Legislature this year approved its own moratorium bill, but Hochul’s office described the legislation as complex and said it needed additional work. Instead, the governor opted for an executive order that would take effect immediately.

New York, at this stage, has not been a destination for the biggest hyperscale data centers.

Keep reading

AI Companies Absorbing Office Space At Record Pace: Report

Artificial intelligence (AI) firms are absorbing office space in primary markets such as San Francisco and New York City at a record pace, and the sector’s voracious demand for office space to build out development teams and products has begun spilling into a select subset of submarkets as well.

National AI office demand was up 85 percent in the 12 months through May and spiked 179 percent in major AI hubs, a new AI report published on July 9 by AI-powered commercial real estate platform VTS states.

AI companies represented office demand of 16.8 million square feet across 17 markets during the period, VTS senior research manager Rene Moreira noted.

However, three metro areas represented nearly two-thirds of total office demand from AI companies, with San Francisco—the global epicenter for AI talent and development—accounting for 25 percent.

Office properties in San Francisco and Silicon Valley, California, and New York accounted for 63 percent of all current AI leasing, Moreira said.

“San Francisco alone sits at 5 million square feet, nearly a third of the national total,” he said.

Unprecedented office demand from AI companies in San Francisco is powering the city’s office market to a modest recovery after the COVID-19 pandemic. In the second quarter of 2019, San Francisco’s office market hit a vacancy rate of 4.7 percent. Vacancy soared following work-from-home initiatives, however, reaching 30 percent in 2023 and topping out at 35.7 percent as recently as the second quarter of 2025, the City of San Francisco reported.

San Francisco’s office vacancy stood at 32.6 percent at the end of the first quarter of this year.

“San Francisco’s 81 active AI requirements average 62,000 square feet, 2.3 times the average tech requirement across all markets,” Moreira said.

The 45 active AI office lease searches in New York average 61,00 square feet each, while the 58 active searches in Silicon Valley average about 48,000 square feet, or 2.8 million square feet of office space.

Each submarket caters to different AI users, VTS noted. San Francisco is the headquarters of AI pioneers Anthropic (Claude) and OpenAI (ChatGPT), while Silicon Valley’s AI firms tend to be chip designers, hardware manufacturers, and infrastructure providers. New York’s AI companies are skewed toward enterprise-level AI firms, a nod to the city’s massive financial, legal, and media industries. AI firms in Washington, such as Anduril, Palantir, and Shield AI, serve the defense industry.

Keep reading

Pollution from Musk’s unpermitted xAI power project hits hardest in Black communities

Elon Musk’s artificial intelligence company xAI has installed 59 natural gas turbines for its Colossus 2 data center project in Tennessee without securing federal clean air permits, according to communications between regulators and xAI representatives.

Potential emissions from the turbines are far beyond the threshold that would require a federal permit, and would be released near predominantly Black communities already estimated to be suffering disproportionately high rates of lung disease, according to a Reuters analysis based on government data and information in the correspondence with regulators.

The findings, which have not been previously reported, reflect how exploding electricity demand from AI data centers is driving companies to build off-grid power plants at a pace outstripping environmental oversight, with potentially big risks to public health.

The number of unpermitted turbines identified by Reuters is about double what xAI has publicly acknowledged. The company previously said it was running 27 unpermitted turbines for Colossus 2 as of January and has argued the permits are not required. At least 57 of the 59 turbines are located in Mississippi, just over the state line from Tennessee where the data center is located.

The xAI turbines are among scores of off-grid power plants for data centers proposed or under construction around the country. Local authorities often fast-track approvals in just weeks or months, without the years of environmental studies and public hearings typically required for such power generation projects that connect to the grid, Reuters has reported.

Keep reading

REPORT: Russia Intercepts Western AI Drones Smuggled for Deep Strikes on Military Bases

Russian security forces have seized a cache of Western-manufactured, artificial intelligence-guided drones that were allegedly smuggled into the country for planned attacks on remote military aviation bases in the Ural Mountains and the Far East.

The Federal Security Service (FSB) announced Monday that operatives intercepted the hardware before it could reach the Shagol air base in the Chelyabinsk region and the Ukrainka air base in the Amur region. Several individuals involved in the transit operation have been detained.

Covert Smuggling Operation

According to the FSB, the drones first entered Russian territory using balloons and larger fixed-wing unmanned aircraft, landing in the Bryansk border region. Smugglers then concealed the devices inside specially modified trailers equipped with false bottoms, transporting them across the country disguised as shipments of household appliances, reported Sputnik.

AI-Guided Strike Capabilities

Russian authorities stated that the confiscated drones originated from the United States, Britain, Canada, and Sweden. Each unit was armed with more than one kilogram of explosives and featured advanced AI navigation systems engineered to evade Russian electronic warfare and signal-jamming networks.

Parallel to Ukrainian Tactics

The intercepted plot bears similarities to a 2025 Ukrainian operation known as “Spider’s Web,” in which trucks with retractable wooden roofs were used to transport and launch drones. That campaign reportedly damaged around 20 aircraft and successfully struck the Ukrainka base.

In recent months, Kyiv has intensified long-range strikes on Russian defense enterprises and energy infrastructure located deep inside the country. These attacks have taken major oil refineries offline, contributing to widespread fuel shortages.

Kremlin Accuses West of EscalationI

n response to the expanding reach of Ukrainian deep-strike operations, Moscow has sharpened its accusations against Western nations, claiming they are directly enabling the attacks.

Last week, Kremlin spokesman Dmitry Peskov declared that Western support for Ukraine had transformed Russia’s “special military operation” into a full-scale war.

Keep reading