Fedi nerds (including yours truly): I wouldn't let that slop anywhere near my hobby open source project
DOD: As planning for a potential strike in Iran was underway, Maven, powered by Claude, suggested hundreds of targets, issued precise location coordinates, and prioritized those targets according to importance … The AI tools also evaluate a strike after it is initiated
😬
https://wapo.st/4rcM6dC
(email walled)
Three years into the AI hype wave, it keeps happening:
1) AI vendor tries to make BS machine look more reliable by having it link sources
2) BS machine BSes the sources
"…its “sources” linked to spammy copies of legit websites, or other archived copies that aren’t the actual source page. Some sources even went to completely unrelated links that weren’t written by the person whose work they were supposedly an example of"
https://www.theverge.com/ai-artificial-intelligence/890921/grammarly-ai-expert-reviews
With a little help from a Ouija board, even the dead ones can opt out, I suppose. I mean, it's as much their voice as the Grammarly version
"you can create a DLP policy to help protect against the use of sensitive information types (SIT), such as credit card numbers, passport identification, or social security numbers in Microsoft Copilot 365 prompts" - So hypothetically, if one were to include random but formally valid SSN or CC values hidden in your emails or documents, would it stop users of this feature from using their microslop on it? 🤔
https://learn.microsoft.com/en-us/purview/dlp-microsoft365-copilot-location-learn-about
You can use Microsoft Purview Data Loss Prevention (DLP) targeted at the Microsoft 365 Copilot and Copilot Chat location to help prevent the use of sensitive information types in prompts and files and emails that have sensitivity labels in Microsoft 365 Copilot and Copilot Chat prompts.
In today's #AIIsGoingGreat (ht @platypus*), Microsoft security does a nice writeup of SEO bros abusing "summarize with AI" buttons to inject "memories" into AI assistants… and then goes on to offer mitigations like "be sure to hover links before you click them" and "regularly check your AI memories" … because if the last 30 years of infosec has taught us anything, it's that user vigilance is the first and best line of defense, right?
https://www.microsoft.com/en-us/security/blog/2026/02/10/ai-recommendation-poisoning/
That helpful “Summarize with AI” button? It might be secretly manipulating what your AI recommends. Microsoft security researchers have discovered a growing trend of AI memory poisoning attacks used for promotional purposes, a technique we call AI Recommendation Poisoning.
Bonus #AIIsGoingGreat: After assuring us recent incidents were only coincidentally connected to AI*, Amazon "summoned a large group of engineers to a meeting on Tuesday for a “deep dive” into a spate of outages, including incidents tied to the use of AI coding tools"
RE: https://mstdn.social/@rysiek/116211625230754185
Not that humans are immune to screwing up TZ/DST logic of course, but I feel like the odds the offending logic was Claude vomit are pretty high, and the fact the RFO doesn't address this is pretty telling
Also this is a good illustration of why the "buT It WRiTes WorKInG CodE" argument is fairly unpersuasive on its own
https://mastodon.social/@rysiek@mstdn.social/116211625348719630
LOL. But will there be any reflection on how it got this far? Did no one stand up and point out the many obvious reasons was likely to be a total shit-show, or were they ignored?
I mean, one of their examples is "where is the closest public bathroom that isn’t completely disgusting" and what are the odds Google's LLM has accurate, up to date information about this? (and if, in fact, google does have realtime surveillance of public restrooms, I may have a privacy-related followup)
https://www.theverge.com/tech/893262/google-maps-gemini-ai-ask-maps-immersive-navigation
RE: https://mathstodon.xyz/@mjd/116224397839379268
This will be an interesting test of the AI companies fine print "don't use this great amazing world transforming genius machine for anything serious, lol" disclaimer.
My totally uneducated IANAL guess is OpenAI will win, if they don't settle to make it all go away. As much as the disclaimers are obvious CYA, OpenAI hasn't (AFAIK) explicitly promoted it ChatGPT for litigation
https://mastodon.social/@mjd@mathstodon.xyz/116224398083939471
Original pro se case https://www.courtlistener.com/docket/69634076/dela-torre-v-nippon-life-insurance-company-of-america/ which as far as I can tell seems to have effectively ended with the plaintiff agreeing to arbitration and not being sanctioned into oblivion
Case against OpenAI
https://www.courtlistener.com/docket/72365583/nippon-life-insurance-company-of-america-v-openai-foundation/
RE: https://infosec.exchange/@josephcox/116256386324754543
Shot: "Kantor told 404 Media that artificial intelligence is writing more than half the app’s code these days"
Chaser:
https://mastodon.social/@josephcox@infosec.exchange/116256386410352613
I do wonder though, does anyone involved actually want it or think it will work, or does it exist purely for management have have an ✨AI story?
https://www.404media.co/tinder-plans-to-let-ai-scan-your-camera-roll/
"The leak, which Meta confirmed, happened when an employee asked for guidance on an engineering problem on an internal forum. An AI agent responded with a solution, which the employee implemented – causing a large amount of sensitive user and company data to be exposed to its engineers for two hours" - I would like to see a description of what happened not filtered through generalist press…
#AI takes another journalism job: https://www.theguardian.com/technology/2026/mar/20/mediahuis-suspends-senior-journalist-over-ai-generated-quotes
🤔 #AGI… Artificial Gross Incompetence? Another Grift Industry? A Giant Inferno?
Grammarly dude is very quick to say their "expert review" impersonation feature was a bad feature created by a small team that he wasn't involved with and how they already killed before the lawsuit was filed, but very much less to keen to concede that attributing slop to people without permission or compensation is just a shitty thing to do
Who decided to (mis)appropriate CQ from hams when "Slop Overflow" was right there?!
(reposting in the correct thread)
So I thought this was SBU trolling, given the very strong indications Ukrainian drones *have* gone through Belarusian airspace in the attacks mentioned, but on closer inspection, it appears Google decided to translate "ДПСУ" as "State Border Guard Service of the Republic of Belarus" 😬 . Oddly, translating just the paragraph gets it right. This seems like a very #LLM-y failure mode
Also the earlier story about AI translations sure sounds familiar: "implemented new policies and restricted a number of contributors who were paid to use AI to translate existing Wikipedia articles into other languages after they discovered these AI translations added AI “hallucinations,” or errors, to the resulting article"
https://www.404media.co/ai-translations-are-adding-hallucinations-to-wikipedia-articles/
#AIIsGoingGreat… so great it's opening up a whole new branch of psychiatry: "There seem to be three common delusions in the cases Brisson has encountered. The most frequent is the belief that they have created the first conscious AI. The second is a conviction that they have stumbled upon a major breakthrough in their field of work or interest and are going to make millions. The third relates to spirituality and the belief that they are speaking directly to God"
https://www.theguardian.com/lifeandstyle/2026/mar/26/ai-chatbot-users-lives-wrecked-by-delusion
Bonus #AIIsGoingGreat (HT @davidgerard*) I was joking about collateralized GPU obligations**, but Quinn Emanuel Urquhart & Sullivan, LLP ain't: "This note surveys the major financing mechanics—direct loans, SPV structures, securitizations, and GPU-collateralized facilities—and identifies nine categories of emerging litigation risk"
https://www.quinnemanuel.com/the-firm/publications/client-alert-emerging-litigation-risks-in-financing-ai-data-centers-boom/
* https://mastodon.social/@davidgerard@circumstances.run/116320136755099444
** https://mastodon.social/@reedmideke/115080228685202839
Yet more #AIIsGoingGreat: Dude sets a slop bot loose on wikipedia and then lectures them about being "constructive", despite unapproved bots being prohibited by longstanding policy "They probably should have used this more as a learning experience because this type of AI agent interaction is about to become the new normal, and they will need more constructive ways of working with them"
More (HT @nora*) "The worst thing that happens is he gets banned and something gets taken down. The best thing that happens is he actually contributes some useful things to Wikipedia … I was real surprised by the reaction, because I just kind of assumed that there [were] a lot of agents that were potentially contributing to Wikipedia at this point" - Dude just incapable of understanding that other people might not want to clean up his shit
* https://mastodon.social/@[email protected]e/116319649387305289
Today's #AIISGoingGreat is amazing, revolutionary technology, created at vast expense, which will reshape the global economy and virtually everything we do. Get on board or be left behind*
https://www.microsoft.com/en-us/microsoft-copilot/for-individuals/termsofuse
* for entertainment purposes only
(HT @mhoye https://mastodon.social/@mhoye@cosocial.ca/116325416135307490)
Uh, as bad as it is, I'm not totally convinced spilling PII to the ad tracking industry is the worst thing in this picture 'Perplexity responds to a basic prompt like “What is the best treatment for liver cancer?” by volunteering that “I can help you interpret a specific scan report, biopsy result, or proposed treatment plan if you share more details”' 😬
https://arstechnica.com/tech-policy/2026/04/perplexitys-incognito-mode-is-a-sham-lawsuit-says/
If we ever recover our senses, there's still gonna be like a K-Pg boundary of radioactive slop that will haunt us for years
https://mastodon.social/@neurovagrant@masto.deoan.org/116341969357555709
#AIIsGoingGreat "Before a hearing, [federal judge] Rodriguez might also ask AI to suggest questions to ask an attorney or identify weaknesses in a plaintiff’s argument. In an area of law in which he feels particularly well-versed, Rodriguez sometimes — after deciding on his judgment — uses AI to draft the ruling he will issue’"
"Rodriguez … raised the example of reviewing a motion for summary judgment, where a party asks a court to resolve a case before trial. Those motions can be accompanied by volumes of evidence, such as deposition transcripts, that can be time-consuming to sort through … I’m uploading everything … And then I’ll ask, ‘Identify any potential statements made in this age discrimination case that appear discriminatory.’" - JFC, sure seems like a lotta ways that could go wrong
LOL, OpenAI shells out "low hundreds of millions of dollars" to buy TBPN, says TBPN will "help with marketing and communications at OpenAI but keep their editorial independence"
Today's #AIIsGoingGreat: Major LLM chatbots pick up a very fake disease from very fake preprints and do not reliably identify it as fake. OpenAI claims this is NBD because… "The models that power today’s version of ChatGPT are significantly better at providing safe, accurate medical information, and studies conducted before GPT-5 reflect capabilities that users would not encounter today"
Useful #LLM poisoning tip: "The format of the fake-disease experiment — and the way the results pretended to be from an official source, namely an academic paper, might have been a key factor in its success. [another study] found that LLMs are more prone to hallucinate and elaborate on misinformation when the text they’re processing looks professionally medical — formatted like a hospital discharge note or clinical paper — than when it comes from social-media posts"
Today's #AIIsGoingGreat (HT @ai6yr*) 'the chatbot said it was not trying to replace my physician; the outputs were for educational purposes. “Think of me as a med school professor, not your doctor” … The bot said the best way to get an interpretation of my health data was just to “dump the raw data,” like clinical lab reports, and tell it what my goals were' - Makes Microsoft's "copilot is for entertainment only" look positively responsible!
https://www.wired.com/story/metas-new-ai-asked-for-my-raw-health-data-and-gave-me-terrible-advice/
* https://mastodon.social/@[email protected].org/116388641691454041
Superintelligence, of which YOU SHOULD NOT RELY UPON OUTPUTS FOR ANY PURPOSE
¯\_(ツ)_/¯
First person to prompt-inject virtual Zuck into giving them a billion dollar bonus gets a billion dollars!
Oh, what's that, Meta doesn't trust virtual Zuck to handle real money? Why is that, too menial for a superintelligence? 🤔
RE: https://esq.social/@D_J_Nathanson/116420731934704356
Today's #AIIsGoinGreat: It's easy to imagine a product like this "fixing" the hallucinated citation problem by using non-AI code to look up citations in Westlaw's database, and nagging the model to fix ones that don't check out. Which will get you valid citations, but unfortunately for Westlaw and the #ChatGPTLawyer in this case, verifying the citation actually supports the thing it's cited for is an entirely different and much harder problem
https://mastodon.social/@D_J_Nathanson@esq.social/116420732051053223
I on the other hand predict that a great deal of entertainment will ensue!
(for outsiders watching the trainwreck unfold)
https://mastodon.social/@gamingonlinux/116453942800480300
"To reach that bare minimum of 7 percent, Gartner forecasts that large AI companies would need to earn cumulatively close to $7 trillion in AI-driven revenue through 2029" - It's OK, I'm sure some banner ads and horny chatbots will cover it
Google's search for in-the-wild prompt injection involved some regexes and… feeding the content to an #LLM? 🤨 "These candidates were then processed by Gemini to classify the intent of the suspicious text, and to understand whether they were part of the overall document narrative or suspiciously out of place"
and no, they do not discuss whether Gemini was successfully prompt-injected by the any of the content it examined
More seriously, they assess that most of what they found was jokey and/or low sophistication but there's no discussion of whether they encountered anything likely to succeed against commonly used AI tools
https://security.googleblog.com/2026/04/ai-threats-in-wild-current-state-of.html
Today's #AIIsGoingGreat (HT everyone) is of course the sloperator who vibe-deleted their prod database… and also somehow took at face value the "confession" of the bot which (per their story, at least) deleted their database. Notably they quote the "confession" verbatim but not the prompt that triggered it
"frontier models fail to accurately predict their own token usage (with weak-to-moderate correlations, up to 0.39) and systematically underestimate real token costs" - Approaching parity with human programmers' cost/schedule estimation!
The wide adoption of AI agents in complex human workflows is driving rapid growth in LLM token consumption. When agents are deployed on tasks that require a significant amount of tokens, three questions naturally arise: (1) Where do AI agents spend the tokens? (2) Which models are more token-efficient? and (3) Can agents predict their token usage before task execution? In this paper, we present the first systematic study of token consumption patterns in agentic coding tasks. We analyze trajectories from eight frontier LLMs on SWE-bench Verified and evaluate models' ability to predict their own token costs before task execution. We find that: (1) agentic tasks are uniquely expensive, consuming 1000x more tokens than code reasoning and code chat, with input tokens rather than output tokens driving the overall cost; (2) token usage is highly variable and inherently stochastic: runs on the same task can differ by up to 30x in total tokens, and higher token usage does not translate into higher accuracy; instead, accuracy often peaks at intermediate cost and saturates at higher costs; (3) models vary substantially in token efficiency: on the same tasks, Kimi-K2 and Claude-Sonnet-4.5, on average, consume over 1.5 million more tokens than GPT-5; (4) task difficulty rated by human experts only weakly aligns with actual token costs, revealing a fundamental gap between human-perceived complexity and the computational effort agents actually expend; and (5) frontier models fail to accurately predict their own token usage (with weak-to-moderate correlations, up to 0.39) and systematically underestimate real token costs. Our study offers new insights into the economics of AI agents and can inspire future research in this direction.
A technology so compelling and transformative it needs a billionaire-backed astroturf campaign to promote it
#AIIsGoingGreat who could have predicted that an always-on sycophantic delusion machine could send vulnerable people into delusional spirals?
"The White House has asked a group of tech companies to answer a set of questions this week about how to ward off digital attacks that frontier artificial intelligence tools could soon enable" - which sounds like an invitation to burn taxpayer billions on AI cyber snake oil but… "The four people said some industry representatives were confused by the questions they received, several of which were seen as vague"
https://www.politico.com/news/2026/04/30/white-house-ai-cyber-threats-mythos-00902045
#AIIsGoingGreat "Banks are hunting for new ways to offload risks tied to a glut of data centre debt as the race to build AI infrastructure stretches financing limits among the largest global lenders … Lenders, including JPMorgan and MUFG, have spent more than six months distributing $38bn of construction debt tied to a data centre project leased to Oracle in Texas and Wisconsin"
https://www.ft.com/content/08aba5e4-5834-4e79-a48d-989a2c5bad0f
#AIIsGoingGreat "An acclaimed Canadian fiddle player has launched a $1.5m civil lawsuit against Google, alleging that the online giant defamed him by falsely identifying him as a sex offender in an AI-generated summary of his life and career… he had learned of the inaccurate information when the Sipekne’katik First Nation cancelled a concert appearance planned for 19 December, after members of the public complained, citing the misinformation they read on Google"
Also:
1) The Sipekne’katik First Nation later issued a public apology to MacIsaac, saying: “Decisions were based on incorrect information generated through an AI-assisted search, which mistakenly associated you with offenses unrelated to you. We deeply regret the harm this caused to your reputation and livelihood.”
2) MacIsaac’s lawsuit alleges that Google had never contacted him or offered an apology over the error
Given the fairly irrefutable concrete harm involved, I predict google will settle, and the "can tech megacorps be held liable for the things their BS machines say" question will be kicked down the road again
RE: https://infosec.exchange/@agreenberg/116533336872355044
Democratizing Software Development™ is going great
https://mastodon.social/@agreenberg@infosec.exchange/116533336935148460
Reed Mideke
