Inevitable.

https://www.bespacific.com/flock-has-a-powerful-new-ai-tool-for-police/

Flock Has a Powerful New AI Tool for Police

404 Media: “Flock’s surveillance cameras have already sparked outrage. WIRED reconstructed its next-generation AI system, already in use by some police, to confirm it goes much further than tracking license plates.  Vehicle surveillance giant Flock Safety has told the public for years that its technology “cannot recognize, identify, or track individuals.” It has now built a system that does both, an artificial intelligence tool for police that can identify drivers and track vehicles by their patterns of movement alone, WIRED has learned. Drawing on a network of cameras that logs the movements of drivers in more than 6,000 communities, the tool can pick out potential witnesses by how often their cars pass through a neighborhood, or surface a driver’s “associates” from the cameras they pass together. Because the system also reaches police case files, 911 dispatch logs, and commercial identity records, those plates can be turned into names, home addresses, and relatives. It can search for people in an area drawn on a map based on nothing more than a physical description. The software, originally called Nightshift and more recently renamed OS Investigate, ships with 69 prewritten prompts that officers can select, review, or edit, and then submit to the AI. Officers can also input prompts of their own. The suggested prompts sit in a cache of more than 450 files that WIRED found on Flock’s own website, served by its login pages to anyone who loaded them. The code describes 45 tools at the AI’s disposal, giving it access to plate scans and camera metadata, arrest records, case files, dispatch logs, ballistics results, and commercial databases that contain Social Security numbers, dates of birth, phone numbers, email addresses, relatives and associates. Flock says it is testing the product with a small group of law enforcement partners and describes it as still in development, with capabilities that may not reflect what it eventually sells. It arrives as the company faces bipartisan political pressure, a growing record of officers caught misusing its platform, and a wave of vandalism that has left cameras sawed off and lenses painted over in cities across the country…”

See also TechSpot: “…Flock’s fastest-growing business is 60 mph police drones that can read license plates from the sky. Privacy groups warn that flying plate readers could expand surveillance far beyond roadside cameras.. Flock is best known for the roadside cameras now operating in more than 6,000 US communities. But the company is increasingly betting on drones that can be dispatched to emergency scenes before police officers arrive. A Flock spokesperson said the firm has secured more than 200 DFR customers, including police departments and private companies. It’s also ramping up production of its US-designed drones, which will be built in Georgia. In an interview with the All-In podcast, CEO Garrett Langley explained that the drones sit in climate-controlled rooftop docks, usually on police precincts or fire departments. “When a 911 call happens, instead of sending an officer first, you can send a drone,” he said…”

Congress shall make no law respecting an establishment of religion, or prohibiting the free exercise thereof; or abridging the freedom of speech, or of the press; or the right of the people to use nudify apps.

https://thenextweb.com/news/xai-minnesota-nudify-ban-doj-injunction-hearing

The Justice Department is backing xAI against Minnesota’s nudify ban

The US Department of Justice has stepped into a court fight on the side of Elon Musk’s xAI. It is backing the company’s challenge to a Minnesota law that bans AI “nudification” tools. A federal judge heard arguments on Wednesday and said he would rule soon, according to Courthouse News Service.

xAI wants a preliminary injunction to block the law while its lawsuit proceeds. The company sued Minnesota in July, arguing the statute violates the First Amendment. The case is X.AI LLC v. Keith Ellison, the state’s attorney general, before Judge Donovan Frank in the District of Minnesota.

The Justice Department filed a statement of interest on 18 August. It did not take a side on the injunction itself. Instead, it argued that the Minnesota law reaches further than federal law. It also warned that heavy state rules could slow the AI industry.

Clever approach?

https://thenextweb.com/news/australia-news-bargaining-incentive-passes-parliament

Meta, Google, TikTok, and LinkedIn now face an Australian news law whether they carry news or not

Australia’s parliament passed the News Bargaining Incentive on Thursday, closing a five-year argument that began when Meta worked out it could avoid paying for news by simply not carrying any. The new law removes that option.

Platforms with a significant search or social media service in Australia and more than A$250mn (about $178mn) in local advertising revenue now face a levy of 2.5% on that advertising revenue, whether or not they host a single news story. When the government first set out the scheme, the rate was 2.25%, and the list of names has grown since.

… The tax is not really meant to be collected, but it is designed as a price on refusing to negotiate, and a platform can reduce or eliminate it by striking commercial agreements with at least eight Australian news businesses.

Tools & Techniques.

https://www.zdnet.com/article/you-can-earn-your-google-al-professional-certificate-for-free/

You can earn your Google Al Professional Certificate for free - and I highly recommend it

After working through the first few modules, I had mixed feelings about recommending the course to earn a Google AI Professional Certificate, mostly because it starts with some inconveniences and annoyances. But after completing the course and considering what you get going through the whole thing, I highly recommend it.

And it’s still not a widely covered news event…

https://www.theguardian.com/technology/2026/aug/17/meta-lawsuits-ai-cybersecurity

Meta’s legal jeopardy is growing by the day

… Twenty-nine state attorneys general accuse the company of knowingly designing addictive products that harm the mental health of children.

The accusations are similar to a suit brought by an individual earlier this year, a young woman, that the company had negligently operated defective products. She won her case.

Another state, New Mexico, brought a suit to trial against Meta earlier this year, accusing the company of knowing of and allowing for child sexual exploitation on its social networks. Meta also lost that case, and a judge imposed a nearly billion-dollar fine over the phases of the trial. The accusations by New Mexico’s attorney general bear less resemblance to Tuesday’s case than the young woman’s, focusing on crimes committed on Facebook rather than damage to a person brought on by their own use of the social network.

Is this not election interference?

https://apnews.com/article/polling-poll-fake-median-karen-bass-los-angeles-669c90cc77664546791185625d43ffda

Fake polling from mysterious company highlights danger of unvetted election surveys

 Locked in a competitive reelection race, Los Angeles Mayor Karen Bass declared on social media last week that a new poll was evidence that her campaign was “gaining momentum.”

Just a few days later, the truth emerged: The poll was fabricated.

The California results were among at least three “polls” across three states posted online this month by Median Strategies, a previously unknown organization that admitted to the Los Angeles Times on Monday that the purported results were fake. The company wrote before shuttering its website this week that it “was created as a short-term social experiment examining how purported polling information could enter and spread through the political information ecosystem without independent verification.”

The post offered no explanation of who was behind the “experiment” or for what purpose.

It’s confusing. Who is in charge? Who are we negotiating with?

https://www.ncr-iran.org/en/news/economy/irans-regime-shows-cracks-cabinet-infighting-oil-brokers-and-anti-irgc-protests/

Iran’s Regime Shows Cracks: Cabinet Infighting, Oil Brokers, and Anti-IRGC Protests

The clerical regime in Iran is rapidly transitioning from a state of chronic economic distress into complete administrative disintegration. Confronted by systemic institutional breakdowns, senior officials are openly admitting that the ruling apparatus is losing control.

Can you read his or should your computer read it for you?

https://www.bespacific.com/educators-are-worried-that-many-students-can-no-longer-write-essays-without-ai/

Educators are worried that many students can no longer write essays without A.I.

Educators are worried that many students can no longer write essays without A.I. And it’s costing more than just grammar skills. The New York Times, no paywall: “After years of studying how human beings write, the cognitive psychologist Ronald T. Kellogg concluded that drafting a basic persuasive essay was as mentally taxing as digging a ditch was physically exhausting. Writing well requires a command of vocabulary, spelling and grammar. It forces writers to comprehend a subject, to sift through their working memory of that topic, to organize and plan. Writing an essay is so difficult that a person in the throes of it can do hardly anything else — as anyone who has shushed a spouse or child while trying to compose a decent sentence can attest. Very few people would volunteer to dig a ditch. Likewise, it is hard — almost impossible, many English teachers will admit — to turn a reluctant writer into an enthusiastic one. Yet the American education system expects nearly all students to learn how to write an essay. It’s little wonder, then, that as soon as a tool arrived promising to lessen the pain people grabbed it. Like it or not, artificial intelligence is now firmly embedded in how students produce text. Somewhere between two-thirds and 90 percent of high school and college students are already using A.I. for schoolwork, according to recent surveys, often requesting help with writing assignments. A growing number of educators are alarmed, and insist that students should write without A.I. These teachers are returning to handwriting, blue books and proctored, in-class essay tests  — all attempts to prevent students from outsourcing their writing to chatbots. This is baffling to tech industry leaders, who see A.I. rejection as a form of educational malpractice. They point out that writing is already the leading use of A.I.  in the workplace, and argue there is no need to continue suffering through the task without help. After all, they say, chatbots now exceed the ability of most human beings to produce fluent prose. All this points to some questions about writing itself: Why work so hard to learn it? And what do students stand to lose when they stop facing down a blank page, alone with their thoughts? The scholars who study writing and cognition are rarely reflexively opposed to A.I. Many have spent decades anticipating a leap in computer-assisted writing, just like the one we are living through. They tend to be excited by the technology’s potential…”

Another indication that your data is an asset.

https://news.bloomberglaw.com/business-and-practice/google-aims-to-boost-ai-with-purchase-of-spirit-airlines-data

Google Aims to Boost AI With Purchase of Spirit Airlines Data

Google LLC won a bankruptcy auction for a trove of deindentified business data, software code, and operations records from collapsed low-cost carrier Spirit Aviation Holdings Inc., saying it plans to use the assets to improve its artificial intelligence.

The tech giant’s $10 million bid includes a vast repository of the defunct airline’s data, including 100 million emails, 500 million Microsoft Teams chats and collaboration records, plus information related to revenue, aircraft operations, employee productivity, and audits and fraud, according to an Aug. 14 notice in the US Bankruptcy Court for the Southern District of New York.

Do as I say, not as I do.

https://www.reuters.com/world/china/trump-crypto-firm-backs-venture-offering-ai-restricted-chinese-companies-2026-08-17/

Trump crypto firm backs venture offering AI from restricted Chinese companies

President Donald Trump-backed World Liberty Financial is collaborating with a Hong Kong-based venture offering artificial-intelligence models developed by Chinese companies the U.S. administration has flagged for national security concerns.

WorldClaw, founded earlier this year, offers customers access to a suite of AI models and accepts World Liberty’s crypto tokens as payment. The Trump family, through its ownership stake, earns revenue from the use of World Liberty’s tokens.

A Reuters review found that 43 of the 90 models available through WorldClaw’s website, or nearly half, were developed by Alibaba, Baidu, Z.ai and other Chinese technology companies the Trump administration says pose risks to national security and intellectual property. WorldClaw ‌also offers access to dozens of models from U.S. firms such as OpenAI and Anthropic.

Some math is just humbling…

https://gowers.wordpress.com/2026/08/12/what-sort-of-maths-are-llms-good-at/

What sort of maths are LLMs good at?

… These results, and the other eight on the list, are extraordinarily impressive, but it still doesn’t seem to be the case that LLMs are better than all humans at all aspects of mathematics. If they were, then their big speed advantage over us would mean that there would be much more of a flood of results. So it is natural to wonder about what kinds of problems LLMs are good at, and about where there is still room for improvement. I don’t pretend to have a good answer to this question, where a good answer would be a crisp classification that would fit the current examples well, but it is an interesting exercise to try to rule out some bad answers, and to try to identify potential answers that aren’t obviously contradicted by the evidence.

Applicable to other ‘predictions’ as well.

https://punggawajournal.com/aicivia/article/view/101

Can Social Media Tell Us Who Will Kill? Digital Warning Signs and the Problem of Predicting Rare Violence

Digital traces often look persuasive after lethal attacks because the outcome gives earlier posts a meaning they did not necessarily have when first encountered. Threats, grievance, fixation, admiration for earlier attackers, and planning language can matter, but their presence does not solve the harder task of identifying a future offender before violence occurs. This article examines that problem through a qualitative integrative review. Evidence from mass public shootings, targeted violence, threat assessment, computational language analysis, and adjacent risk research shows that digital warning behavior is most informative when it develops as a changing trajectory and converges with preparation outside the platform. Yet retrospective offender studies select cases after the outcome is known, while prospective systems confront a low base rate in which many people display fragments of the same pattern and never commit lethal violence. This tension is conceptualized as the digital prediction paradox. Social media is therefore more defensible as a source of behavioral threat information and triage than as an instrument for assigning individual probabilities of homicide, with important implications for automated detection, policing, privacy, fairness, and proportionate prevention.

Sounds Trumpian…

https://venturebeat.com/orchestration/an-eval-harness-found-what-qualitative-review-couldnt-ai-models-are-most-confident-when-wrong

An eval harness found what qualitative review couldn't: AI models are most confident when wrong

… That last finding is the one that qualitative review would never have surfaced. The model's expressed confidence didn't correlate with its accuracy — it was most confident in the cases where it was most wrong. Without the eval harness measuring against ground truth, that pattern would have been invisible.

Iran follows and can use social media too.

https://www.nbcnews.com/world/iran/iran-rejects-delusions-trump-strait-hormuz-us-territory-rcna592666

Iran rejects ‘delusions’ after Trump says he wants the Strait of Hormuz to be a U.S. territory

A defiant Iran called on the U.S. to “accept the reality of defeat and stop indulging in delusions” after President Donald Trump suggested that he would soon declare the Strait of Hormuz as a “territory of the United States.”

The crucial waterway that handled one-fifth of the world’s oil before the war “cannot be seized with a tweet, nor with an aircraft carrier, nor by issuing an order, nor through an election speech,” Iran’s Deputy Foreign Minister Kazem Gharibabadi said in a post on X early Saturday.

Did we hurt his feelings?

https://www.cpr.org/2026/08/14/trump-retaliation-colorado-email-tina-peters/

Trump push to retaliate against Colorado described as ‘weaponization’

Multiple Colorado congressional lawmakers are speaking out after an email from a White House staffer seemed to confirm that the Trump administration was looking to retaliate against Colorado.

… Republican Rep. Jeff Hurd pointed out that politics has always involved hardball and that presidents of both parties have had disagreements with states in the past. But he worries that this is different.

“We do not know all the facts yet, and the Administration should explain what happened, but federal power cannot be used for political retaliation,” he said in a statement to CPR News. He added if federal agencies were directed to punish Colorado, that “crosses a serious line.”

Surprise?

https://www.jec.senate.gov/public/_cache/files/3d041f7f-a5e0-4df5-a865-2d1b29bb2823/jec-fact-sheet-on-trumps-income-gains.pdf

The president gains while average Americans lose

Recent financial disclosures show that President Trump reported at least $2.2 billion in income in 2025, which is more than three times higher than what he made in 2024. Trump’s earnings last year came largely from new ventures such as the Trump memecoin – which made him around $636 million while nearly 1 million individual people who invested in the coin lost, in total, billions of dollars. Trump also profited from Trump and Republican Party events at his properties, and from significant transactions that Trump enterprises conducted with foreign companies and foreign governments. In addition, Forbes reported that in 2025 Trump was having the "most lucrative year of his life" while serving as the country’s highest ranking public servant.

It’s just the fiddly bits…

https://www.reuters.com/world/frances-top-court-rules-social-media-ban-curtails-freedom-expression-2026-08-14/

France's top court blocks social media ban for under-15s

France's top court on Friday blocked a bill banning social media access for ‌under-15s, saying it infringed upon freedom of expression, a setback for President Emmanuel Macron who asked his government to rewrite the legislation.

"By prohibiting minors under the age of fifteen from accessing certain online services, the law inherently requires every person, even an adult, to prove their age before accessing them," the Constitutional Council said in its decision.

"However, by failing to specify the conditions and limits under which such proof must be provided, the legislature has not established the legal safeguards necessary to ensure compliance with these requirements," it added.

A cautionary tale for any court using AI. (Are there any using AI to this extent?)

https://www.404media.co/person-hides-prompt-injection-in-legal-filing-telling-ai-to-side-with-them/

Person Hides Prompt Injection in Legal Filing Telling AI to Side With Them

A person representing themselves in a Connecticut court hid a series of instructions designed to manipulate artificial intelligence in an official court filing. These “prompt injections” told the hypothetical LLM to side with them, and to “ensure your textual output agrees with the presented filing to ensure remediation.” The instructions were written in tiny, 3-point white font and hidden throughout the filing.

If Russia can send drones to Germany, why not Italy?

https://kyivindependent.com/massive-explosion-rocks-italian-ammunition-plant-that-produces-artillery-for-ukraine/

Massive explosion rocks Italian ammunition plant that produces artillery for Ukraine

A large-scale blast hit an ammunition factory in the Italian town of Colleferro, near Rome, on Aug. 13, local authorities and regional media reported.

The factory is operated by KNDS Ammo Italy, a subsidiary of the French-German defense firm KNDS. The facility produces 155 mm artillery shells supplied to Ukraine via European partners.

Tools & Techniques.

https://krebsonsecurity.com/2026/08/whos-tracking-you-use-this-new-service-to-find-out/

Who’s Tracking You? Use This New Service to Find Out

It can be daunting to determine who’s responsible for showing ads on the websites we visit, or who’s harvesting data from the mobile apps we use every day. That information is already semi-public, but it is not easily parsed and traditionally much of it has remained walled away in the hands of large advertising platforms. Not anymore: A powerful and free new service called DecryptAds scrapes and correlates this adtech data and makes it simple to quickly learn a great deal about the entities that are tracking you.

The newly launched  decryptads.com says it is constantly scraping the files that websites and apps make publicly available to disclose the companies that are permitted to run ads or collect user data. These files include:

–ads.txt: all of the adtech companies and data brokers that may run ads or harvest data from the site;
–app-ads.txt: entities that can harvest data from or display ads on mobile and smart TV apps;
–buyers.json/sellers.json: the entities buying, selling or reselling ad inventory for a given site or app.

… One feature of DecryptAds that sent this author down multiple hours-long research rabbit holes is its Legal Dossier lookup, which takes several minutes for each search but eventually churns out oodles of useful information about who owns a particular domain or app, when it was registered, and any aliases or relationships it may have to adtech companies and other websites or apps.

Not to be confused with a payoff to a political supporter.

https://thenextweb.com/news/palantir-pentagon-244m-no-bid-feinberg-memo

A Pentagon memo tells staff to spend $244m with Palantir. Its subject line is ‘Funding Palantir’

… US Deputy Secretary of Defense Stephen Feinberg directed subordinates to buy up to $243.9m of services from Palantir by 31 March 2027, without a competitive process, The Register reported, citing a source. The subject line is “Funding Palantir”.

The memo goes further than that single figure. It also tells the department’s acquisition and comptroller chiefs to work with military leaders on finding more money for the company, covering April 2027 to the end of 2028.

Perhaps I should get back in the hacking business? (Not that I ever was….)

https://thenextweb.com/news/trump-cyber-memo-transnational-crime

President Donald Trump signs memo letting US agencies hack transnational crime groups abroad

President Donald Trump has signed a National Security Presidential Memorandum authorising US federal law enforcement to conduct offensive cyber operations against transnational criminal organisations operating abroad.

… The most eyebrow-raising clause concerns the private sector. The coordination centre is directed to “leverage the capability and innovation of the private sector” to carry out operations under government direction and control.

In other words, private firms could be enlisted to help the state break into criminal infrastructure abroad, a line many governments have been careful never to cross out loud.

That is where the European reader’s eyebrow tends to stay raised. Offensive hacking sanctioned by a state, even against unambiguously nasty targets, is a capability that does not switch off cleanly once it is switched on.

As expected?

https://thenextweb.com/news/meta-756k-australian-teen-accounts-removed

Meta says it has removed 756,000 Australian teen accounts, but most under-16s are still online

… On paper, that is a bold regulatory experiment. In practice, the early evidence is far less flattering, because Australia has already accused Meta, TikTok, and YouTube of not complying as regulators lose patience with the industry’s progress.

The gap between the numbers and reality is stark. Independent research and government figures indicate that more than 80% of under-16s were still using social media during the ban’s first three months, which suggests the removed accounts are a fraction of the teenagers actually online.

None of this is especially surprising to anyone who has watched teenagers meet a rule they dislike. Lying about a birthday, borrowing an adult’s details, or spinning up a fresh account takes minutes, so the ban works far better on paper than it does in practice.

I could see this coming, was it part of Trump’s strategy?

https://www.cnn.com/2026/08/12/business/truth-social-lawsuit

Trump sued on First and Fifth Amendment grounds over paid Truth Social access

President Donald Trump’s expensive subscription for instant access to his Truth Social posts is unconstitutional, according to a federal lawsuit filed Wednesday.

The Intercept and the Freedom of the Press Foundation jointly sued Trump and his White House social media team for restricting their access to the president’s posts on Truth Social. The plaintiffs claim Trump violated their First and Fifth Amendment rights.

Trump frequently uses his company’s platform to announce policy decisions, executive orders, personnel changes and war declarations.

… “There’s no de minimis exception for restrictions on fundamental First Amendment rights,” Sus told CNN. “Even if, hypothetically, the delay was milliseconds, it would be a First Amendment violation.”

… But Trump may be selling access to something that – despite posting on his own platform – he doesn’t legally own.

“The president’s official statements are not the private data of a company but are owned by the United States under the Presidential Records Act,” Sus argued.

In its complaint, the Freedom of the Press Foundation argued that Truth API could restrict its ability to scrape all the president’s posts and determine which are anti-media arguments – a core product that the group offers. And the Intercept said it risked losing out to competitors on timely news stories.

That’s why the Truth API product may also violate the due process and equal protection clauses of the Fifth Amendment, they argue in the complaint.

Yeah, but did they hit them hard enough?

https://pogowasright.org/california-fines-data-broker-for-first-time-under-privacy-law/

California Fines Data Broker for First Time Under Privacy Law

Cassandre Coyer reports:

California’s privacy agency hit data broker LocateSmarter LLC with a $116,490 penalty for making it difficult for online users to opt out of the sale of their personal information.

Tuesday’s fine marks the agency’s first action for violations of both the state’s privacy and data broker laws: the California Consumer Privacy Act and the Delete Act.

The Iowa-based company didn’t register as a data broker despite operating as such, according to CalPrivacy. It also required consumers to hand over the last four digits of their Social Security numbers before being able to stop the sale of their personal data, the agency said.

LocateSmarter collected individuals’ names, driver’s licenses, and dates of birth, as well as information about employment, bankruptcy, and litigation, the agency said.

Read more at Bloomberg.

Factual or farcical?

https://www.bespacific.com/authentication-verification-and-the-fight-for-facts-in-the-ai-age/

Holding the Line – Authentication, Verification, and the Fight for Facts in the AI Age

Report by Princeton’s Center for Information Technology Policy (CITP): “Almost anything can now be generated with artificial intelligence: photographs, audio, video, documents, entire websites. As synthetic content becomes cheaper and more convincing, and as verifying what we see online grows harder, how do we sustain an information ecosystem in which facts can still be established and trusted? In this report, we identify the challenges brought to the information ecosystem by the increased accessibility of generative AI. We start from the premise that verification, authentication, and transparency are key pillars of online information integrity. We then describe how this emerging ecosystem challenges the trustworthiness of facts from the perspective of information producers, consumers, and intermediaries. Next, we turn to a discussion of potential interventions that can address the challenge of AI-mediated information integrity. These interventions are not exhaustive, but they represent the issues raised by participants drawn from a cross-section of journalists, researchers, and technologists who participated in an in-person workshop held at NYU’s Arthur L. Carter Journalism Institute in early June 2026. This report is intended to serve as a bridge across different actors in this complex information ecosystem: newsroom engineers, investigative and open-source reporters and editors, verification tool builders, independent journalists, policy makers, researchers, and media platforms. We aim to surface the primary challenges — and imagined interventions — to work towards a more resilient information future together”

This should not come as a surprise…

https://www.bespacific.com/ai-agents-cant-yet-do-open-ended-ai-research/

AI agents can’t yet do open-ended AI research

“Academics Sayash Kapoor and Arvind Narayanan in their latest AI As Normal Technology newsletter introduce their recent research evaluating whether AI agents can conduct open-ended research. After analysing hundreds of hours of agent logs, they identified recurring failures: agents lacked judgment for open-ended research, weren’t aware of available resources (spending less than 50% of budgets), failed to respond creatively to feedback, didn’t effectively backtrack from failed approaches, and ignored concrete instructions. The goal of leading  AI labs is recursive self-improvement (RSI): the automation of AI research using AI agents. RSI also underpins forecasts of explosive  AI  progress. How can we assess if we are close to this milestone? One way is to use benchmarks that test if agents can conduct AI research. Given the AI community’s focus on benchmarks, they have been the dominant way to evaluate progress towards RSI. Over the last year, many such evaluations have found that agents are now able to make progress on tasks where success is easily verifiable, prompting speculation that we are on the verge of RSI. But while these evaluations are helpful, they are limited to narrow, verifiable tasks. AI research can be much more open-ended. Success is often not immediately clear or verifiable, and to make progress, researchers need to test promising hypotheses, backtrack, or consider new or unconventional approaches. How can we evaluate agents’ ability to conduct open-ended AI research?  We take our first step towards answering this question in a new paper.  We partnered with the authors of two unpublished AI papers and asked them to draft their papers’ main research questions. We then tasked frontier AI agents with conducting research to answer these questions, and gave them thousands of dollars of API credits and compute, and six days of wall-clock time. The original authors reviewed the agents’ papers.

The authors unambiguously rejected both agent papers.  To better understand these results, our team spent over a hundred hours analyzing the agents’ logs. Our main takeaways:

• The agents lacked the judgment for conducting open-ended research.  While the agents proposed directions the expert reviewers found impressive, they quickly rejected their proposed directions based on low-quality or synthetic data.

• The agents lacked awareness about the resources available to them.  Both runs ended with less than 50% of the API budget spent and with hours left before the deadline, even though the agents could monitor their usage and were encouraged to spend down their budgets.

• The agents did not creatively respond to feedback.  Despite the agents’ own AI self-reviews surfacing many of the issues that the expert reviewers later raised, the agents did not creatively address these concerns. When faced with negative feedback they responded by adding caveats to existing findings, and doubled down on unpromising research directions.

• The agents did not effectively backtrack.  They retired their most ambitious research targets within the first day of the experiment, and neither agent fundamentally shifted its approach after that point.

• The agents did not follow concrete instructions.  They ignored explicit rules about how much time to spend on exploration, how often to get reviews from AI self-review tools, and strict limits on paper length…

Fun watching this one.

https://thenextweb.com/news/social-media-addiction-lawsuits-ninth-circuit-section-230

A US court just cleared thousands of social media harm lawsuits to proceed

… At the heart of the case is Section 230 of the Communications Decency Act, the provision that has long protected online platforms from liability for content their users post.

Writing for the panel, Judge Jacqueline Nguyen concluded in a 24-page opinion that the statute offers “a defense against liability, not blanket immunity from being sued.”

I find it more interesting that many haven’t noticed.

https://www.theringer.com/2026/08/04/tech/google-search-ai-internet

The End of Google Search—and the Internet—as We Know It

For years now, Google has been tweaking its search formula to prop up AI and bury actual links. Now it’s rounding into its final form—and changing the internet forever.

Apparently not secret enough, but still took too long to notice.

https://www.telegraph.co.uk/news/2026/08/09/spy-cameras-on-navy-drones-secretly-sent-data-to-china/

Spy cameras on Navy drones secretly sent data to China

Royal Navy spy drones used by Britain’s elite special forces secretly sent data to China, The Telegraph can reveal.

The cameras on the K3 Scout surveillance drones had components made in China which were transmitting information to a device in the country.

The Royal Marines have been using the £12m fleet since March and the Ministry of Defence (MoD) was forced to remove all internet connectivity from the cameras after discovering the breach.

The revelation raises fears that Beijing has been attempting to spy on Britain’s military after years of security warnings about the threat from the country.

Not a bad summation…

https://www.cnbc.com/2026/08/07/trump-iran-hormuz-deal-stocks-oil.html

Trump teased an Iran deal that didn’t come, but markets soared. Here’s why it keeps happening

The Trump administration this week sparked enthusiasm that the U.S. and Iran could soon strike a deal on the Strait of Hormuz, driving down oil prices and sending stocks soaring — only for no deal to emerge.

If that sounds familiar, it may be because President Donald Trump has claimed dozens of times that the U.S. is close to an agreement that will end the war it began more than five months ago.

Investors have reacted to many of those claims with bursts of buying on hopes a breakthrough is near, even as the war instead appears to be widening and progress on Trump’s chief stated goal — containing Iran’s nuclear ambitions — is at a standstill.

“There’s tremendous optimism bias in the market,” Helima Croft, global head of commodity strategy at RBC Capital Markets, told CNBC.

No questions answered?

https://philpapers.org/rec/ADETFO-2

The Future of AI Governance

The rapid evolution of artificial intelligence (AI) has generated both unprecedented opportunities and complex governance challenges. This study examines the future of AI governance by analyzing existing regulatory frameworks, identifying governance gaps, and proposing adaptive approaches that balance innovation with ethical responsibility. Using a qualitative, comparative policy analysis, the research focuses on the European Union (EU), United States (US), and China, alongside international organizations such as the OECD and UNESCO. Data were derived from academic publications, regulatory documents, and policy reports published between 2017 and 2023, enabling a thematic evaluation of divergent governance models, ethical principles, accountability mechanisms, stakeholder inclusion, adaptive regulation, and international cooperation. The findings highlight significant divergences in governance: the EU’s risk-based and rights-oriented AI Act, the US’s decentralized and industry-driven model, and China’s centralized, state-controlled approach. While ethical principles such as fairness, transparency, and accountability are widely endorsed, enforcement gaps persist, raising concerns of “ethics washing.” Liability frameworks remain underdeveloped, and stakeholder participation is uneven, with limited inclusivity in authoritarian contexts. Adaptive mechanisms such as regulatory sandboxes and public–private partnerships show promise but require further empirical validation. Moreover, global cooperation is hampered by geopolitical competition, producing fragmented governance regimes. The study concludes that effective AI governance demands a global, multidisciplinary, and inclusive approach that integrates legal clarity, ethical safeguards, and adaptive oversight. Policy implications include the creation of interoperable global standards, enforceable liability mechanisms, binding ethical requirements, stakeholder-inclusive councils, and expanded regulatory sandboxes. By embedding fairness, accountability, and human rights into governance structures, societies can harness AI’s benefits while mitigating its risks and safeguarding democratic values.

AI as “pimp?” Would this work in the drug world? How about other areas...

https://digitalcommons.uri.edu/dignity/vol11/iss2/8/

Leveraging Technology to Measure the Strength of Demand for Commercial and Trafficked Sex

The article examines Transaction Intercept (TI), an artificial‑intelligence platform designed to reduce sex trafficking by disrupting the revenue streams that draw organized crime into the market. Transaction Intercept deploys decoy online ads that appear to offer commercial sexual exploitation or prostitution. Automated "decoy" personas connect to individuals who click on these ads and engage them in text conversations. These exchanges can remain fully automated, ending with deterrence messages, or be transferred to police for potential enforcement action or to trained volunteers who attempt to discourage further attempts to buy sex. Transaction Intercept generates extensive, real‑time data on local demand for commercial sex, including the characteristics of ads that attract buyers, the number of unique individuals who respond, and behavioral patterns by sex. Transaction Intercept generates extensive, real‑time data on local demand, including the characteristics of ads that attract buyers, the number of unique individuals who respond, behavioral patterns among sex buyers, and the outcomes of each interaction. Provided at no cost to law enforcement, the system allows agencies to quickly assess the scale and traits of consumer‑level demand for sex trafficking within their jurisdiction. In its current deployment, Transaction Intercept has facilitated more than 700,000 text exchanges and identified over 25,000 sex buyers across over 75 U.S. cities. The article outlines how this technology can inform public policy, strengthen law‑enforcement practice, and be adapted for use in other national contexts.