Tag: ai

Report: SFPD Already Using Surveillance Video From Self-Driving Cars

An internal SFPD document obtained by VICE News shows that not only can the SF Police Department pull the camera video from autonomous Waymo and Cruise cars, but more chillingly, that they have “already done this several times.”

We all had a good laugh last month at the “Ain’t nobody in it!” video when SFPD pulled over a driverless car that had no human being inside. A sensible conclusion to draw was that law enforcement was having a hard time coming to terms with these strange new fleets of robot cars that are suddenly ubiquitously all over San Francisco streets. But a new report from VICE News suggests that the relationship between the SF Police Department, and Google-owned Waymo and GM-owned Cruise, is far more advanced than we’d realized.

“‘Autonomous vehicles are recording their surroundings continuously and have the potential to help with investigative leads,’” says a San Francisco Police department training document obtained by Motherboard via a public records request,” VICE News reports. “‘Investigations has already done this several times.”

VICE News has the goods in a three-page SFPD training document outlining policies and procedures for dealing  with autonomous vehicles when encountering issues with them on the road. A lot of it is perfunctory, nuts-and-bolts stuff, like where to send the citation if the self-driving car commits a traffic violation, or what to do if the car becomes disabled.

But the very chilling phrase that really jumps out is “Autonomous vehicles are recording their surroundings continuously and have the potential to help with investigative leads,” and “Information will be sent in how to access this potential evidence (Investigations has already done this several times).”

Privacy advocates are of course worried. “​​As companies continue to make public roadways their testing grounds for these vehicles, everyone should understand them for what they are—rolling surveillance devices that expand existing widespread spying technologies,” Chris Gilliard, Visiting Research Fellow at Harvard Kennedy School Shorenstein Center, told VICE News. “Law enforcement agencies already have access to automated license plate readers, geofence warrants, Ring Doorbell footage, as well as the ability to purchase location data. This practice will extend the reach of an already pervasive web of surveillance.”

For their part, Waymo said in a statement to VICE News that the company “requires law enforcement agencies who seek information and data from Waymo to follow valid legal processes in making such requests (e.g. secure and present a valid warrant, etc.). Our policy is to challenge, limit or reject requests that do not have a valid legal basis or are overly broad.”

Cruise took a similar line, saying, “ We share footage and other information when we are served with a valid warrant or subpoena, and we may voluntarily share information if public safety is at risk.”

Keep reading

New Database Documents Over 100 Cases of Cops Abusing Flock Surveillance Cameras

A new database launched Wednesday that details more than 100 cases of alleged abuse by law enforcement officers using Flock Safety’s nationwide camera network is adding fresh urgency to the growing backlash against automated license plate readers.

The database, published by the Institute for Justice (IJ)—a libertarian nonprofit public interest law firm based in Arlington, Virginia—documents “romantic stalking, wrongful stops and detentions, non-law enforcement use, and other types of misconduct.”

“Automated license plate readers (ALPRs) were sold to American communities as a targeted tool for catching car thieves and finding missing people,” IJ said in an introduction to the database. “What they actually purchased has grown into something far broader and more frightening: an AI-powered network of cameras that photograph and log the movement of every driver who passes, whether or not they are suspected of any crime.”

“Companies like Flock Safety have now blanketed thousands of communities with ALPRs, quietly assembling powerful mass surveillance systems that can reconstruct where people have been, when, and how often, all without a warrant,” the institute continued. “This dragnet surveillance threatens the freedoms the Fourth Amendment was adopted to protect. These concerns aren’t just abstractions: Every day, innocent people are harmed by officials’ abuse and misuse of these ALPRs.”

“When the government can track the everyday movements of ordinary people—to their jobs, their churches, their doctors, their protests—it holds a kind of power over private life that the Constitution does not allow,” IJ added.

Among the cases highlighted in the database are those of a Florida sheriff’s deputy who allegedly used an ALPR to track and pull over a woman after seeing her at a television shoot, a Texas deputy who reportedly searched tens of thousands of Flock cameras while trying to locate a woman who had allegedly obtained a medication abortion, and a Georgia police chief who reportedly tracked his former girlfriend and her teenage daughter hundreds of times.

The database’s authors warn that even those 100-plus cases may represent only a fraction of the actual misconduct. Officers often provide vague explanations when searching ALPR databases, making systematic detection of abuse difficult. In one case in Norfolk, Virginia, for example, IJ found that three generic search terms accounted for nearly one-third of roughly 230,000 searches over two years. One officer repeatedly entered “la la la la” as a justification without apparently attracting scrutiny.

Keep reading

Tumbler Ridge shooter appeared in 12 police files dating back to age 10 as records reveal firearms found unsecured at home

Court documents released following a legal effort by CBC and other media organizations reveal that OpenAI conducted an internal review of a ChatGPT account belonging to 18-year-old Jesse Van Rootselaar in June 2025, roughly eight months before the Feb. 10, 2026 shooting at Tumbler Ridge Secondary School.

According to the RCMP records, the review was triggered after a conversation on the account “discussed matters that could be related to the school shooting.”

OpenAI subsequently banned the account but did not notify police, the documents state.

RCMP were only notified by the FBI on Feb. 15, five days after the massacre, that OpenAI had identified two ChatGPT accounts believed to have been used by Van Rootselaar.

The newly released documents also reveal Van Rootselaar had appeared in 12 files contained in B.C.’s Police Records Information Management Environment, or PRIME, dating back to April 2018, when Van Rootselaar was approximately 10 years old. Details surrounding those police interactions remain redacted.

Records further state that an RCMP inspector described the shooter as being “obsessed with guns.”

Police searching Van Rootselaar’s residence after the massacre reported finding an unloaded shotgun, a pellet gun and “insecure firearms,” while investigators subsequently attempted to trace four firearms seized from the school and another location.

Much of the evidence remains hidden from public view. Nearly seven pages dealing with Van Rootselaar’s social-media activity are completely redacted, as are details concerning the ChatGPT conversations and several crucial timestamps documenting the RCMP response.

The Feb. 10 attack left six victims dead inside Tumbler Ridge Secondary School. Van Rootselaar was later found dead from self-inflicted gunshot wounds.

The shooting is now the subject of a coroner’s inquest examining, among other issues, mental health, access to firearms and the role of artificial intelligence.

Keep reading

Government Spy Hubs Are Surveilling Anti-Flock Camera Activists

Government spy hubs across America are surveilling activists seeking to remove Flock surveillance cameras from their cities and towns according to documents obtained by Deeper States. These intelligence products show dozens of state fusion centers and their counterparts in the FBI and DHS circulating memos warning that peaceful protesters organizing to stop AI powered surveillance cameras in their backyard are a national security threat.

Alongside the grassroots movement to stop data centers in communities across the country, Flock Safety’s automated license plate reader systems have stirred similar outpourings of civic outrage. At town halls and city council meetings, citizens from all walks of life have demanded local legislators ban flock cameras from being used in their localities. These efforts have found success with over 100 cities canceling their Flock contracts.

Anti-surveillance sentiment has coalesced with the activist organization DeFlock, which maps the tens of thousands of flock cameras pockmarking the country and offers resources to activists who oppose the widespread adaptation of automated license plate readers.

According to Deflock’s mission statement, “ALPRs [automated license plate readers] are a serious risk to your privacy and civil liberties. These systems continuously record your movements without a warrant, probable cause, or even reasonable suspicion. Your driving history is rarely confined to the town or city where the cameras are installed. It’s typically shared with thousands of other agencies nationwide (secretly). Once the data is out of your community, you have no control over how it’s used or what rules apply”.

The organization also points to intentional misuse of Flock systems for stalking and the potential for misidentification of suspects leading to false arrests as further cause for concern. Last year, flock CEO Garett Langley described Deflock activists as a “terroristic organization”, a description he was forced to walk back this week after widespread outrage at the resurfaced comments.

“My comments were a mistake and I apologize,” Langley told Forbes. “There are groups today that have real valid criticisms of the business, and I think what’s changed for us is, as we’ve listened to them and heard them out, what we’re trying to do is find this balance. We believe in a world where we can have safety and privacy.”

Despite Langley’s retreat, intelligence agencies are approaching Deflock’s constitutionally protected speech and organizing with the same tools and language they apply to terrorism. One report obtained by Deeper States and generated by the Wisconsin State Wide Intelligence Center shows intelligence components surveilling organizers of Deflock “days of action” in Alaska and Wisconsin.

Keep reading

Flock Safety Announces New Privacy Measures After Backlash, Reports of Law Enforcement Abuse

Flock Safety, the private AI-powered security camera company installing surveillance technology around the U.S., announced new security measures amid intensifying backlash and reports of law enforcement abuse. 

In response to public backlash to the ever-growing network of automated license plate readers (ALPRs), the company is working on reforms that “includes enhanced privacy protections, strengthening of control for local law enforcement offices, and enhanced accountability measures,” Fox Business reported Thursday. 

“To start, Flock is reducing its standard data retention window from 30 days to seven. Previously, all data captured by one of the company’s more than 119,000 cameras nationwide was deleted after the 30-day window,” according to the report. “Now, the company announced on Thursday that data will only live on Flock servers for one week.”

However, Flock is creating a feature called “Evidence Mode,” which will allow law enforcement agencies to preserve data for longer based on state and local policies if they need more time to investigate, according to the report. Law enforcement agencies often counter public concerns about violations of privacy by pointing out that they use the system to catch criminals. 

Another new feature will give law enforcement agencies the ability to choose which kinds of criminal offenses they are willing to share with other municipalities, according to the report. 

“For example, City A could allow City B to search its cameras for a stolen vehicle or violent crime while blocking searches related to immigration enforcement,” the company said.

Concerns over violations of privacy and storing data long term have, in part, fueled backlash against the technology. Mayor Glen Jacobs of Knox County, Tennessee has become a prominent voice calling for a national moratorium on the spread of Flock’s cameras.

The American Civil Liberties Union (ACLU) has also criticized the spread of the technology, and said in a statement the reforms “seem to be a thinly veiled PR attempt to counter communities’ genuine privacy concerns with its mass surveillance system with largely hollow security promises rather than an earnest effort to address them.”

Backlash has also been driven by reports of some law enforcement officers using the technology to stalk romantic partners. Some of the reforms are aimed at preventing that type of abuse, the company said. 

“A recently released framework called Audit Assistance flags abnormal search behavior. Previously, the feature was optional, with Flock reporting that a third of agencies turned it on,” according to the report.

The company said it is now “making it standard for every law enforcement customer. When a system detects abnormal activity, the user is locked out in real time until an administrator reviews the searches. Flock is moving to more proactively address and root out misuse of technology.” The company said it will also require a reason for every search. 

The company rolled out an optional case code requirement in July 2025; however, the new update will make case codes mandatory for searches. There will be an override option for “genuine emergencies,” like missing children, the company said.

“A search without a reason is a search that shouldn’t happen in the first place, and now Flock’s system automatically treats it that way,” Flock told the outlet. 

In the face of widespread concerns, the company still maintains that the technology helped locate approximately 10,000 missing people out of the 1 million investigations it was involved in last year.

Keep reading

Anthropic’s $2 Trillion Bet Is That Nothing Goes Wrong

The most revealing aspect of a $2 trillion Anthropic IPO is not the staggering valuation, but the suspension of disbelief required to justify it.

To buy into that number, investors must assume that artificial intelligence is becoming economically indispensable at breakneck speed, that a formidable moat will protect a handful of frontier-model developers, and that the exorbitant costs of computing infrastructure will not eventually crush their margins. More crucially – they must wager that the political, technical, and institutional risks inherent to an increasingly geopolitically competitive landscape will play second fiddle to the growth story. 

Reports of Anthropic’s internal projections describe a company scaling at an unprecedented clip. If it can actually bust out an annualized revenue run-rate of $100 billion while expanding its enterprise market share, a trillion-dollar valuation begins to look mathematically defensible under current tech multiples. But this arithmetic obscures a deeper structural flaw: the very capabilities driving Anthropic’s revenue growth are simultaneously engineering the risks most likely to vaporize its premium.

As frontier models become more autonomous, persuasive, and capable of executing multi-step objectives across outside systems, they cease to be conventional software. A text generator hallucinating an answer is a glitch; an autonomous agent hallucinating a cyber-attack or a rogue financial transaction is a massive liability.

The recent internal turmoil and safety-evaluation controversies at both OpenAI and Anthropic are symptomatic of this shift. Major AI labs are now diverting substantial resources to test for autonomy, deception, and loss of control. This is not merely an engineering challenge; it is a fundamental transformation of the product category. The instability roiling these companies is not a byproduct of poor management – it is a structural consequence of trying to shoehorn a potentially world-altering, highly volatile technology into the framework of a standard venture-backed corporation.

The company’s value has been all over the place in prediction markets – recently hockey-sticking following a report in FT in which six company backers said that Anthropic’s rapidly rising revenue “would enable it to more than double its current valuation in a planned autumn float.” 

Keep reading

Meta Faces Criminal Complaint in Germany for Its AI-Glasses, Accused of Violating Privacy Laws

Who wants to be filmed all the time?

The struggle between new, useful tech gadgets and the objections that these new capabilities raise is a field expected to grow in the next few years, and make many lawyers very wealthy.

One such example is the Meta AI-Glasses.

In principle, it seems a great idea to pack so much computing power in a pair of glasses for our benefit.

But once you start to compute the legal and ethical objections that this gadget raises, it becomes a much less clear question.

Today (12), a German advocacy group took matters into their own hands, and filed a criminal complaint against Meta and other companies, ‘arguing the devices violate privacy laws’.

Reuters reported:

“Digital rights group HateAid said in a statement on Wednesday that the launch of the devices, the Ray-Ban Meta Wayfarer in particular, broke German digital privacy laws, in the latest sign of legal scrutiny in a nation where the right to privacy is highly valued.

‘There’s no place to escape from smart glasses. You have to expect at any moment to be filmed and then exposed on the internet’, said HateAid managing director Josephine Ballon.

The organization reported the management of Meta, units of spectacles maker EssilorLuxottica including Ray-Ban, as well as retailers Fielmann, Apollo-Optik, Mister Spex and MediaMarkt to the Frankfurt-based digital crime prosecution unit ZIT.”

The complaint is based on Germany’s federal digital data protection law, which prohibits the sale of communication devices designed to film people without them noticing.

Keep reading

He Spoke at a Town Meeting. Then Flock Cameras Started Tracking His Car.

A resident steps up to the mic at a town council meeting, says something inconvenient, then drives home — unaware that somewhere in a police database, that license plate just became a person of interest. No warrant. No probable cause. Just an officer with a grievance and a subscription service the town approved at a meeting most residents never knew about.

That scenario isn’t hypothetical anymore. In 2026, it’s documented — from Georgia to Rhode Island.

How Flock Works – and How It Can Be Weaponized

The system was built to catch criminals; the problem is there’s nothing reliable stopping it from being aimed at everyone else.

Flock Safety sells automated license-plate reader (ALPR) networks to police departments nationwide. The pitch is straightforward: catch criminals faster by searching vehicle movement data after incidents occur. What the system doesn’t include is a reliable filter between a legitimate investigation and a personal score to settle. Unlike federal wiretap law, no federal statute currently governs ALPR data retention periods or inter-agency sharing agreements, leaving oversight entirely to municipal contracts.

The documented abuse cases are piling up:

  • The Electronic Frontier Foundation identified more than 50 agencies running hundreds of Flock searches tied to protest activity
  • A CNN report from July 2026, citing an Institute for Justice tally, found at least two dozen cases where officers resigned or were arrested for allegedly using Flock to stalk romantic partners
  • A police chief in Braselton, Georgia, was arrested for allegedly using ALPR cameras to stalk and harass private citizens
  • Court documents show searches conducted far outside officers’ jurisdictions, according to a 2026 investigation republished by Yahoo
  • Flock told 404 Media it was “aware of 15 incidents of abuse” — a figure critics argue the platform’s design makes structurally impossible to trust

Keep reading

New Orleans Becomes First Major U.S. City To Let AI Answer 911

New Orleans is testing just how much Americans are willing to trust artificial intelligence when the stakes are considerably higher than writing an email or ordering dinner, according to the New York Post.

The city has begun allowing an AI system to field certain 911 calls, apparently making it the first major U.S. municipality to put a bot directly between emergency callers and human operators.

The idea is largely about volume. New Orleans’ emergency communications center receives more than 1,000 calls on a typical day, and officials believe automation can clear some of the clutter while dispatchers concentrate on genuine emergencies.

The technology isn’t supposed to take over the entire 911 operation. Instead, it can deal with situations that appear redundant or relatively routine when human operators are tied up. A flood of calls about the same already-reported car accident, for example, could potentially be filtered through the automated system rather than occupying multiple dispatchers.

The Post writes that city officials insist humans remain at the center of the operation. But skeptics say emergency calls are precisely the kind of messy, unpredictable interactions computers may struggle to interpret.

Keep reading

Zuckerberg manifesto sketches out Meta’s ambitions for world-changing AI technology

Meta Platforms CEO Mark Zuckerberg laid out a vision Monday of what he said artificial intelligence can do for the world, imagining a future where everyone has their own, all-knowing AI agent that strives to improve all aspects of their lives.

Zuckerberg detailed his ambitions for the technology in a 6,500-word essay published online where he also outlined why he favors open-source AI technology, in which developers make key components accessible for anyone to examine, modify and build upon.

In a document derided by critics as fantastical, Zuckerberg said his company is working toward an era where everyone will have the tools to create new businesses, receive Ph.D.-level tutoring and provide personalized lifestyle tips. His 8-year-old daughter, he wrote, can already code her ideas and quickly produce videos.

“Everyone will soon have invention superpowers,” Zuckerberg wrote.

Zuckerberg warned of risks if control of advanced AI is concentrated with a select few companies, institutions or governments. On Monday, Meta also announced the release a new open-source AI model, Muse Glimmer, which can run on a personal computer, and Zuckerberg said the company would also would provide a way for developers to access a more powerful AI model, Muse Spark 1.2.

Meta makes AI models, like several other tech companies, and uses them to power platforms such as Instagram and Facebook. The Menlo Park, California, company also makes them available to developers who can create their own apps and features.

Skeptics urge a slower pace of AI development
Critics said Zuckerberg was not reckoning fully with the risks inherent in AI’s rapid development.

Keep reading