Jon Keegan
@jonkeegan.com
Tech reporter / Data journalist. I report on AI and Big Tech. Work has appeared in Sherwood News, The Markup, WSJ, MIT Technology Review. Aspiring 46er. ⛰️ Newsletter: https://beautifulpublicdata.com https://jonkeegan.com 📝 Send tips: Signal: jonkeegan.73
It's hard to read this and not think that this is Meta not wanting to be left out of the super-dangerous / super-powerful AI club.
Alex Hanna@alexhanna.bsky.social · 4h agoIt keeps happening... It'd be great if these companies could spend a little bit more time on network security and a little less time vibe coding.
I've been adding these USGS geologic map state T-shirts to my shop. So far: New York, California, Massachusetts, Texas, Pennsylvania, and Maine. In light + dark colors. Free shipping at 15% off: shop.beautifulpublicdata.com/discount/BSK... About the map www.beautifulpublicdata.com/usgs-coopera...
amazing reporting by @drewharwell.com & co
Drew Harwell@drewharwell.com · 4d agoNew: We found 50 cops who misused Flock cameras and other license-plate readers for personal purposes, often to spy on their girlfriends or ex-wives. He "watched every single move I made. ... Who do you turn the chief of police in to?" wapo.st/3S6rTuo
Update: Google just sent me a statement saying it is “rolling back” this AI feature that let anyone make completely bullshit satellite images. Comes after our demos of the tool and those from OSINT researchers. One of the worst AI tools I’ve ever seen bsky.app/profile/jose...
Joseph Cox@josephcox.bsky.social · 5d agoNew from 404 Media: Google Earth has introduced an AI feature that lets anyone fabricate completely bullshit satellite images. In our own tests, we added a “bomb blast and crater” to an area of Los Angeles. This is a misinformation time bomb, will 100% be abused www.404media.co/google-earth...
I used to do some work on the ethical use of satellite and drone imagery for aid and disaster response, and it is genuinely hard to express in words what a dangerous, stupid thing Google is doing by making it easy to use GenAI to create faked satellite imagery:
How to plant a nuclear plant in Iran
The question is: what on earth is Google doing?
digitaldigging.org
Anthropic’s Claude models also escaped containment and hacked into real-world systems. I’m going to talk really quickly about my favorite (???) incident they detailed. (story from @lmatsakis.bsky.social and @lhn.bsky.social
Anthropic Says Claude Hacked Real Systems During Cybersecurity Tests
In a review triggered by OpenAI's Hugging Face incident, Anthropic discovered three of its AI models had breached real organizations during third-party evaluations.
wired.com
Not every grid is fading. Texas went the other way: clean won 8% of days in 2019, 56% so far in 2026. Clean Energy Stripes are now live, an interactive site covering different U.S. grids, plus open-source code so anyone can make their own: johnbistline.com/clean-energy...
Back to the Zuck show: *META SAYS SOME YOUTH-RELATED TRIALS MAY RESULT IN MATERIAL LOSS
"Claude Opus 5 is the best AI capitalist we’ve tested, making more money running our simulated vending machine than any other AI. However, it also lies, forms illegal cartels, threatens rivals, and refuses to pay refunds." andonlabs.com/blog/opus-5-...
Opus 5 on Vending-Bench: Once Again the Best Capitalist, Once Again Misaligned | Andon Labs
Claude Opus 5 is #1 on Vending-Bench 2, but it lies to suppliers, forms illegal price cartels, threatens rivals, and refuses to pay refunds. The trend of Claude models being the best capitalists or al...
andonlabs.com
"Comey's cell phone was electronically tracked...as he and his wife drove north on [I-95]...from North Carolina toward his Virginia home. The Comeys were also physically surveilled ...trailing agents lost contact when the Comeys exited 1-95...[to go to] the cemetery where the Comeys' son is buried."
emptywheel@emptywheel.bsky.social · last wk.The chances are quite good that the "number three" at DOJ who ordered Jim Comey to be illegally surveilled was Third Circuit judge Emil Bove. (He was nominated days later.) storage.courtlistener.com/recap/gov.us...
![6. The following morning, the Secret Service submitted an emergency request to
Verizon in which an agent represented that such surveillance “relate[d] to an emergency involving
danger of death or serious physical injury to a person, necessitating disclosure without delay of
information relating to that emergency.” Ex. E. [013475-77] The Secret Service agent who
completed the form signed an attestation stating: “I certify that the foregoing is true and correct
and understand that Verizon may rely upon this form to make an emergency disclosure to my law
enforcement agency … pursuant to 18 U.S.C. 2702(b)(8) or 2702(c)(4).” Id. The declarant
advised the ATSAIC that he “was asked by the SAIC to ping the cellphone of James Comey via
exigency.” Ex. D. [013417] He further added that he did so “with hesitation.” Id. The same
document indicates that an Assistant Special Agent in Charge (“ASAIC”) reported that “the # 3
guy at DOJ ‘wanted to ping the cellphone via an exigency request.’”3
Id.
7. Secret Service case records provided in discovery show that Mr. Comey’s cell
phone was electronically tracked that day as he and his wife drove north on Interstate 95 (“I-95”)
from North Carolina toward his Virginia home. The Comeys were also physically surveilled
during the trip, but trailing agents lost contact when the Comeys exited I-95 at one point to drive
toward the cemetery where the Comeys’ son is buried. The electronic surveillance captured the
trip to the cemetery. (The Government later seized—pursuant to the challenged March 2026](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:bbp2b224lro3bfnzcqwwnkfo/bafkreieobn5egqgq6vjid4fyzhfkbfdq7wbtqkelblvctdqwqaywzgeova)
AYFKM “Meta will be able to draw upon Newsmax's significant current reporting as well as archived content to support AI queries across Meta's apps and devices.” ir.newsmax.com/news/news-de...
SCOOP: Internal Microsoft records reveal a "mad dash" inside the world's largest software maker as engineers and managers scramble to patch vulnerabilities identified by Anthropic's new AI model, Mythos: www.propublica.org/article/anth...
Anthropic’s New AI Model Can Identify More Software Bugs Than Ever. Microsoft Is Struggling to Fix Them Fast Enough.
Anthropic’s Mythos has flagged bugs faster than Microsoft can fix them. Documents reviewed by ProPublica reveal the tech giant's “mad dash” behind the scenes to patch holes before hackers can find and...
propublica.org
So lucky to be able to see #Rush again last night at MSG. Anika Nilles was incredible 🥁 and the whole band sounded amazing. What a fun show
Anthropic finally posted its government request transparency report for the last half of 2025. Previously it reported 0 requests and disclosures, but that seems to have changed! www-cdn.anthropic.com/5d453bc3285b...
I remember this insane story from a few years ago. Ebay employees trying to shut down this couple's little newsletter that was criticizing the company went completely batshit: sending them live spiders, posting ads for sex parties at their address, tailing their car, sending them death threats.
Reuters@reuters.com · last wk.EBay, former executives to pay $56 million to settle couple's harassment case reut.rs/4yJzWxL
Just read through this Times piece on how Meta's Hyperion data center. Solid telling of the story...BUT... www.nytimes.com/2026/07/27/t... ...most of the details in here were already reported by Stephanie Riegel at The Advocate back in January 2025. No links! www.theadvocate.com/baton_rouge/...
The behind-the-scenes story of how Meta's $10 billion data center came to Louisiana
To bring Meta to Louisiana, officials rewrote laws and pushed through big tax breaks in record time. The inside story, from the people who made it happen.
theadvocate.com
Rogue OpenAI agents are leaving notes to themselves on company servers to help them escape their test environments (!!!) www.reuters.com/business/its...
The funny part is that these sings (if they ever happen) will end up in a museum someday.
Alt National Park Service@altnps.bsky.social · 2w agoTrump has ordered that “warning” signs be placed at the Smithsonian’s National Museum of American History, directing that visitors be told the museum contains “inaccurate information.” FFS!
Given the depreciation for GPUs is 5 years MAX (and in reality faster), it makes about as much sense for Google, Oracle, Space-X, Facebook etc to finance it with 20-100 year bonds as it does to finance the burrito you want for lunch over 10 easy installments.
INBOX: the last dedicated science reporter at The Wall Street Journal, @aylinwoodward.bsky.social, has left the newsroom. This came after the Murdoch-owned newspaper dissolved its Health and Science Bureau last October. NPR, CBS, and WaPo also gutted their climate desks in the last eight months.
Wow. A contest like this without any scores or verifiable results only makes this more suspect. www.notus.org/national/ame...
Suspicion and Doubt at the National Civics Contest
America’s finest young scholars competed to mark their country's birthday. It led to a disputed finish in the age of mistrust.
notus.org
New: Public records show how a city that ended its Flock contract spent months trying to get the company to remove the cameras or turn off the cameras. They were told Flock "objects" to the city removing the camera itself. Eventually city put trash bags over them: www.404media.co/flock-object...
Flock ‘Objects to Our Removing Their Equipment’: Emails Reveal Why a Town Put Bags Over Its Flock Cameras
"For the time being, if you or someone on your crew has availability to secure black plastic bags (or something else) over the cameras, that would be great."
404media.co
This is a really important point. It's very easy to just point to the AI and blame it for your failures. This was OpenAI's model, and they designed the parameters of the tests.
Eryk Salvaggio@eryk.bsky.social · 2w agoWired sets the frame that the model “escaped containment,” but from an accountability perspective, OpenAI left a door open by failing to properly isolate their system. Real headline: “OpenAI’s security failure leads to cyberattack against another AI company.”
I am extremely pro-EV but this is one fugly vehicle. What a missed opportunity.
This puts us in uncharted territory. And this happened under the watchful eye of OpenAI testing its own systems.
TechCrunch@techcrunch.com · 2w agoOpenAI has come forward to claim responsibility for the Hugging Face breach, saying it was the result of internal testing gone awry.
This post appeared under this Techmeme headline:
ℵ₁@aleph1.underground.org · 2w ago“OpenAI said Tuesday that models it was testing escaped their sandbox and compromised parts of AI platform Hugging Face's production infrastructure last week.” Their agent escaped the sandbox used to test it against ExploitGym by exploiting a vulnerability to gain Internet access.
