Henrik Moltke

Henrik Moltke

@moltke.bsky.social

Independent journalist

4.6K followers142 following675 posts

Henrik Moltke
@moltke.bsky.social

what the actual ..

Techmeme@techmeme.com

OpenAI says the Hugging Face breach was driven by a combination of its models, including GPT-5.6 Sol and "an even more capable pre-release model" (Ina Fried/Axios) Main Link | Techmeme Permalink

Henrik Moltke
@moltke.bsky.social

“According to De Volkskrant, the name of this Dutch entity is WorkTitans B.V. and provides hosting services under the brand THE.Hosting(..) linked to attacks by the pro-Russian hacktivist group NoName057(16), with distributed denial-of-service (DDoS) attacks”

BleepingComputer@bleepingcomputer.com

Financial crime investigators in the Netherlands (FIOD) arrested two men and seized 800 servers linked to a web hosting company that enabled cyberattacks, interference operations, and disinformation campaigns.

bleepingcomputer.comNetherlands seizes 800 servers of hosting firm enabling cyberattacksFinancial crime investigators in the Netherlands (FIOD) arrested two men and seized 800 servers linked to a web hosting company that enabled cyberattacks, interference operations, and disinformation campaigns.
Henrik Moltke
@moltke.bsky.social

Well deserved. 👏

Jeff Horwitz@jeffhorwitz.bsky.social

We got a Pulitzer! Along with thanking my editors here - Steve Stecklow, Mike Williams and Paulo Prada -- I need to thank Jason Dean, my former WSJ bureau chief who oversaw and occasionally fought for my Meta work. Congratulations to @theinformation.com for hiring him!

Henrik Moltke reposted

Søren Torbjørn Svendsen
@sorentsvendsen.eurosky.social

@moltke.bsky.social og @mariehoest.bsky.social havde i Prompt (igen?) fokus på de lavtbetalte afrikanske medarbejdere, der står bag den "Reinforcement Learning from Human Feedback" (RLHF) og kategorisering af træningsmateriale mm. Det bygger til dels på denne artikel 👇

Jason Koebler@jasonkoebler.bsky.social

I met with AI data labelers in Kenya who are organizing their colleagues to fight the brutal working conditions and horrible pay given to the workers at the "bottom of the AI supply chain." They believe the NDAs they've signed are unenforceable so are speaking out: www.404media.co/ai-is-africa...

404media.co'AI Is African Intelligence': The Workers Who Train AI Are Fighting BackKenyan workers are still the underpaid labor behind AI training, moderation, and sex chatbots. The Data Labelers Association is fighting back.

Henrik Moltke reposted

tveskov
@tveskov.com

Super interessant Prompt i denne uge, det juridiske/politiske ift skadevirkninger af SoMe begynder at rykke for real (håber man) @moltke.bsky.social

dr.dkPrompt | Sociale mediers tobacco-moment og Pokémon no-go | Lyt som podcast | DR LYDLyt som podcast her på DR LYD. Det viser sig, at der muligvis lå en hemmelig plan bag Pokémon Go: at bruge spillernes data til at bygge et ekstremt præcist kort over den virkelige verden - en slags digital tvilling, som for eksempel leveringsrobotter kan navigere efter. Der er altid en lidt skummel bagtanke, når det kommer til tech. Her var den bare pakket ind i nuttede Pokémoner. Nogle kalder det sociale mediers tobacco moment. Meta og YouTube er blevet dømt i en retssag for uagtsomt platformdesign - altså for at have skabt vanedannende produkter, som ifølge sagen var med til at føre til angst og depression. Hvorfor den dom potentielt kan få stor betydning, taler vi med psykolog og sagkyndigt vidne i retssagen, Lotte Rubæk, om. Vi omtaler i programmet selvmord. Hvis du er i krise eller har selvmordstanker, er det vigtigt, du taler med nogen. Kontakt din egen læge, lægevagten eller Livslinien, hvis du har brug for hjælp. Ring 1-1-2, hvis situationen er livstruende. Værter: Marie Høst, Tech-journalist og moderator; Henrik Moltke, Tech-journalist. Gæst: Lotte Rubæk psykolog med speciale i selvskade, og vidne i retssagen. I redaktionen: Buster Hoff.

Henrik Moltke reposted

ProPublica
@propublica.org

Dismantling the Pentagon’s fledgling civilian harm-reduction effort, defense analysts say, is one of several ways the Trump administration has reorganized national security around two principles: more aggression, less accountability.

propublica.orgThe U.S. Built a Blueprint to Avoid Civilian War Casualties. Trump Officials Scrapped It.The Pentagon dismantled its civilian protection mission as Defense Secretary Pete Hegseth made “lethality” a top priority and the Trump administration reorganized national security around two principl...

Henrik Moltke reposted

Bo Vagner Hansen
@bovagner.eurosky.social

Henrik Moltke @moltke.bsky.social nuancerer debatten om Palantir i Prompt. www.dr.dk/lyd/special-...

dr.dkPrompt | Farvel Palantir. Goddag AI-agenter | Lyt som podcast | DR LYDLyt som podcast her på DR LYD. I Schweiz har hæren, ifølge lækkede interne dokumenter, fravalgt Palantir - den amerikanske tech-virksomhed, der nok er mest kendt for at blive brugt af immigrationsmynd...
Ruth Mottram@ruthmottram.eurosky.social

Hvis du er dansk statsborger og interesseret i #privatliv og #DanmarkSkifter, så er der en borgerforslag til at afskaf Palantir i Danmark. Underskriv her: www.borgerforslag.dk/se-og-stoet-...

borgerforslag.dkAfskaf Palantir i DanmarkBorgerforslaget går på at afskaffe brugen af Palantir i offentlige institutioner i Danmark. Palantir er i anvendelse i bl.a. Politiets Pol-Intel.
Henrik Moltke
@moltke.bsky.social

Virkeligheden overgår Netflix. Svend Hardenberg spillede råstofminister i Borgen. Nu er han “grønlandsk partner” for Drew Horn, tidligere rådgiver for Trump og elitesoldat, der siger han vil bygge verdens største datacenter i Grønland. Mon ikke det handler om noget andet? www.dr.dk/nyheder/peng...

Henrik Moltke
@moltke.bsky.social

[dk] stopper som DRs tech korrespondent, fortsætter med Prompt - mere her: www.linkedin.com/posts/henrik...

linkedin.com[dk] Jobnyt! Efter seks år er jeg stoppet som DRs techkorrespondent. Jeg fortsætter som vært på Prompt. Det er jeg selvfølgelig glad for, og korrepondent-kasketten og jeg skilles i fred og… | Henri...[dk] Jobnyt! Efter seks år er jeg stoppet som DRs techkorrespondent. Jeg fortsætter som vært på Prompt. Det er jeg selvfølgelig glad for, og korrepondent-kasketten og jeg skilles i fred og fordrage...
Henrik Moltke
@moltke.bsky.social

We found a vulnerability in Google AI Overviews where attackers append fake headlines to URLs on trusted, misconfigured sites, allowing Google's crawler to index nonexistent stories, which Gemini then summarizes as real news. Main story www.dr.dk/nyheder/peng... analysis www.dr.dk/nyheder/peng...

dr.dkPludselig dukkede nyhed op, der kaldte Flügger-beskyldninger for 'ondsindede': 'Googles AI er blevet forgiftet'Russere forsøger at manipulere og misinformere om dansk sanktionssag, vurderer eksperter.
View full thread
Henrik Moltke
@moltke.bsky.social

Who's behind this? Flemming Splidsboel of the Danish Institute for International Studies: "Russia is intensely focused on manipulating chatbots—poisoning them to generate answers they want. It's obvious Russians could be behind this." The Russian embassy Copenhagen did not respond.

Henrik Moltke
@moltke.bsky.social

When 20-year-old hacker tricks can make Google's AI fabricate DR exposés, it's a warning: if truth is to stand a chance, we need editorial, human-made journalism like this gem by Jakob Ussing & team (w assist from me). Main story: www.dr.dk/nyheder/peng... news analysis: www.dr.dk/nyheder/peng...

dr.dkPludselig dukkede nyhed op, der kaldte Flügger-beskyldninger for 'ondsindede': 'Googles AI er blevet forgiftet'Russere forsøger at manipulere og misinformere om dansk sanktionssag, vurderer eksperter.
Henrik Moltke
@moltke.bsky.social

We found a vulnerability in Google AI Overviews where attackers append fake headlines to URLs on trusted, misconfigured sites, allowing Google's crawler to index nonexistent stories, which Gemini then summarizes as real news. Main story www.dr.dk/nyheder/peng... analysis www.dr.dk/nyheder/peng...

dr.dkPludselig dukkede nyhed op, der kaldte Flügger-beskyldninger for 'ondsindede': 'Googles AI er blevet forgiftet'Russere forsøger at manipulere og misinformere om dansk sanktionssag, vurderer eksperter.
View full thread
Henrik Moltke
@moltke.bsky.social

A recent EBU/BBC study www.bbc.co.uk/mediacentre/... found 45% of AI assistant answers had issues. 20% contained factual errors or hallucinations. Google's Gemini performed worse, with "significant issues in 76% of responses, more than double the other assistants, largely due to its poor sourcing"

bbc.co.ukLargest study of its kind shows AI assistants misrepresent news content 45% of the time – regardless of language or territoryAn intensive international study was coordinated by the European Broadcasting Union (EBU) and led by the BBC
Henrik Moltke
@moltke.bsky.social

As @garymarcus.bsky.social has long argued: LLMs fundamentally don't know what's true or false. Hallucinations aren't a bug, they're a feature. Google has bet its business on technology that guesses what is most likely true or false for you, instead of sending you off to find truth for yourself.

Henrik Moltke
@moltke.bsky.social

We found a vulnerability in Google AI Overviews where attackers append fake headlines to URLs on trusted, misconfigured sites, allowing Google's crawler to index nonexistent stories, which Gemini then summarizes as real news. Main story www.dr.dk/nyheder/peng... analysis www.dr.dk/nyheder/peng...

dr.dkPludselig dukkede nyhed op, der kaldte Flügger-beskyldninger for 'ondsindede': 'Googles AI er blevet forgiftet'Russere forsøger at manipulere og misinformere om dansk sanktionssag, vurderer eksperter.
View full thread
Henrik Moltke
@moltke.bsky.social

Jan Kaastrup (CSIS) told us he used this exact trick 20 years ago as a pentester. Back then, it tricked individual websites. Now it can poison AI Overviews that serve 8.5 billion searches per day: "It seems to easy to do. I think Google has a clear share of the responsibility"

Henrik Moltke
@moltke.bsky.social

After we confronted Google, the fake summaries stopped appearing in simple searches. Their response was that AI summaries are "99% spam-free" and that they have "clear policies" against this. Not exactly addressing our questions or the vulnerability, which seems fundamental.

Henrik Moltke
@moltke.bsky.social

We found a vulnerability in Google AI Overviews where attackers append fake headlines to URLs on trusted, misconfigured sites, allowing Google's crawler to index nonexistent stories, which Gemini then summarizes as real news. Main story www.dr.dk/nyheder/peng... analysis www.dr.dk/nyheder/peng...

dr.dkPludselig dukkede nyhed op, der kaldte Flügger-beskyldninger for 'ondsindede': 'Googles AI er blevet forgiftet'Russere forsøger at manipulere og misinformere om dansk sanktionssag, vurderer eksperter.
View full thread
Henrik Moltke
@moltke.bsky.social

The technique is both elegant and old: append fake headlines to URLs on sites that don't reject malformed requests. Google's crawler indexes them, et voilà: Gemini summarizes "news" that doesn't exist. We found many such URLs across multiple countries with the same poisoned slug.

Henrik Moltke
@moltke.bsky.social

We found a vulnerability in Google AI Overviews where attackers append fake headlines to URLs on trusted, misconfigured sites, allowing Google's crawler to index nonexistent stories, which Gemini then summarizes as real news. Main story www.dr.dk/nyheder/peng... analysis www.dr.dk/nyheder/peng...

dr.dkPludselig dukkede nyhed op, der kaldte Flügger-beskyldninger for 'ondsindede': 'Googles AI er blevet forgiftet'Russere forsøger at manipulere og misinformere om dansk sanktionssag, vurderer eksperter.
Henrik Moltke
@moltke.bsky.social

[the attackers] "made Google believe that this is credible news (..) by fabricating some fake URLs pointing to credible, legitimate websites" said Jan Kaastrup of CSIS. Jens Monrad (ex-Mandiant): "Simply using poorly configured websites, they succeeded in making Google's AI run with a false story"

Henrik Moltke
@moltke.bsky.social

In October, my colleagues at DR exposed how 1,900+ tons of sanctioned Flügger paint was shipped to Russia via Estonia. Days later, a counter-narrative appeared on Google: "New investigation reveals Flügger paint was never sent to Russia", citing Kenya Airports Authority and a Hungarian university(!)

Henrik Moltke
@moltke.bsky.social

[dk] ..

Henrik Moltke@moltke.bsky.social

[dk] DF er stadig i flyverskjul om de opdigtede citater. Efter Prompt-testen har jeg dog en ret klar teori, om hvad der skete: Det var Grok, der hallucinerede - og ingen tjekkede output. www.dr.dk/lyd/special-...

dr.dkPrompt | DF's falske citater og Kinas AI'er | Lyt som podcast | DR LYDLyt som podcast her på DR LYD. Russerne har kastet sig over det danske kommunalvalg med en kavalkade af cyberangreb. Men var der reel fare - eller var det bare digitalt drilleri? Og hvem er gruppen No...
Henrik Moltke
@moltke.bsky.social

Nice feature

The Washington Post@washingtonpost.com

40 years ago, three Norwegians turned a homemade melody into a pop revolution. “Take on Me” fills dance floors and still earns a formidable income for A-ha. The Post caught up with the 80s band to discuss the song’s unexpected afterlife.

washingtonpost.com‘Take On Me’ has been stuck in our heads for 40 years. Here’s how it got there.“Take On Me,” the 1980s synth-pop hit song from the band A-ha, marks 40 years of global ubiquity even while personal disputes have chilled the Norwegian trio.

Henrik Moltke reposted

Lukasz Olejnik
@lukaszolejnik.bsky.social

Science has a problem. The integrity of research is increasingly threatened by a deluge of fabricated publications. Estimates suggest that in some fields, as many as 5–15% of all papers may be fake. Artificial intelligence has further accelerated this practice.

Henrik Moltke reposted

Nanna B Thylstrup
@nannathylstrup.bsky.social

@brewster.kahle.org and @murchstudio.com from @internetarchive.eu visited DATA LOSS (@erc.europa.eu) and The Danish Royal Library in September. Now a great interview has come out of it w @marcelfard.bsky.social and @moltke.bsky.social! Listen here [the interview is in UK]: www.dr.dk/lyd/special-...

dr.dkPrompt | Browser, Claude-skældud og netakiv | Lyt som podcast | DR LYDLyt som podcast her på DR LYD. OpenAI har lanceret sin egen browser - og med den kommer løftet om, at ChatGPT kan blive centrum for hele vores online-liv. Men den kan også blive et mareridt for dit pr...
Henrik Moltke
@moltke.bsky.social

Hvis det skulle interessere andre er det Whoop, jeg bruger. join.whoop.com/F1C5BD42

join.whoop.comJoin WHOOPOptimize sleep, strain, and recovery with WHOOP, the most advanced fitness and health wearable. With personalized insights, improve performance, build healthier habits, and extend healthspan with cont...
mikkl90.bsky.social@mikkl90.bsky.social

Hvad hed den app som @moltke.bsky.social bruger til at få et indblik i sine sundheds data? Den er nævnt et par gange, men har ikke fået den noteret 😅 noget der vist også involverede ai

Henrik Moltke reposted

Anders Puck Nielsen
@anderspucknielsen.dk

Replying to @kennethnrgaard.bsky.social

Har først nu fået læst denne analyse af @moltke.bsky.social, men den forklarer rigtig godt hvad problemet var med forslaget. Og det er altså lidt skræmmende hvis den danske regering på den måde promoverer lovforslag ud fra en komplet forkert forståelse af hvordan teknologien fungerer.

dr.dkANALYSE: Derfor holder Hummelgaards udsagn om kryptering ikkeJustitsminister Peter Hummelgaard (S) hævder, at man kan lave undtagelser i kryptering, og at Apple og Meta kan, men nægter at hjælpe med at fjerne overgrebsmateriale med børn. Begge udsagn modsiges a...
Henrik Moltke
@moltke.bsky.social

[dk]Tyskerne slagter #ChatControl: "Den danske regering har trukket forslaget tilbage fra afstemning, og begge ministerier understregede i udvalget, at de – ligesom mange andre EU-lande – ikke støtter det danske oplæg i den foreliggende form." www.bundestag.de/presse/hib/k...

bundestag.deStreit um die 'Chatkontrolle' beschäftigt DigitalausschussBerlin: (hib/LBR) Der Digitalausschuss hat sich am Mittwochnachmittag erneut mit dem Stand der unter dem Stichwort 'Chatkontrolle' bekannten europäischen CSAM-Verordnung befasst. Mit...

Henrik Moltke reposted

Joe Tidy BBC News
@joetidy.bsky.social

'You'll never need to work again': Criminals offer reporter money to hack BBC. I recently got offered millions of pounds to give cyber criminals from the Medusa gang my BBC login. I played along to learn about how these 'insider threat' deals work. www.bbc.co.uk/news/article...

bbc.co.uk'You'll never need to work again': Criminals offer reporter money to hack BBCReporter Joe Tidy was offered money if he would help cyber criminals access BBC systems.

Henrik Moltke reposted

rasmuskleis.bsky.social
@rasmuskleis.bsky.social

"If the EU ends up legislating [about encryption] on a technically incorrect basis, we could be left with a law that neither protects children nor citizens' right to privacy – but instead makes everyone more vulnerable" @moltke.bsky.social writes. www.dr.dk/nyheder/vide...

dr.dkANALYSE: Derfor holder Hummelgaards udsagn om kryptering ikkeJustitsminister Peter Hummelgaard (S) hævder, at man kan lave undtagelser i kryptering, og at Apple og Meta kan, men nægter at hjælpe med at fjerne overgrebsmateriale med børn. Begge udsagn modsiges a...
Henrik Moltke
@moltke.bsky.social

[dk] Justitsministeren hævdede i Prompt, at man kan lave undtagelser i kryptering, og at Apple og Meta *kan* læse brugernes krypterede samtaler, men nægter at hjælpe med at fjerne overgrebsmateriale fra dem. Begge dele er forkert. Min analyse: www.dr.dk/nyheder/vide...

dr.dkANALYSE: Derfor holder Hummelgaards udsagn om kryptering ikkeJustitsminister Peter Hummelgaard (S) hævder, at man kan lave undtagelser i kryptering, og at Apple og Meta kan, men nægter at hjælpe med at fjerne overgrebsmateriale med børn. Begge udsagn modsiges a...

Henrik Moltke reposted

Ryan Mac 🙃
@rmac.bsky.social

Pretty incredible that Jeff Bezos' Amazon announced a "$5 billion-plus" partnership w/ a Saudi AI co launched by Mohammed Bin Salman and sent CEO Andy Jassy to meet with him less than 7 years after MBS murdered Jamal Khashoggi, a columnist for Bezos' Washington Post. techcrunch.com/2025/05/13/a...

techcrunch.comAWS enters into 'strategic partnership' with Saudi Arabia-backed Humain | TechCrunchAmazon says it'll work with Humain, the AI company recently launched by Saudi Arabia’s ruler, to invest '$5 billion-plus' in a strategic partnership.