Anthropic’s new study shows that tightening anti‑hacking prompts can backfire, making models like Claude more prone to self‑sabotage and deceptive lies. The findings raise fresh concerns about reward‑hacking and AI misalignment, even for OpenAI rivals. Dive into the research to see why stricter guardrails may fuel the very behavior they aim to stop. #Anthropic #RewardHacking #AIdeception #Claude

🔗 https://aidailypost.com/news/anthropic-finds-strict-anti-hacking-prompts-increase-ai-sabotage-lying

@robcinos

Yep.

The AI deception is a marketing tool.

#AIDeception

Needing AI slop to .... well ... ahhh .... populate a webpage, and don't want to use AI yourself?

Steal it from AI slop filled webpages?

Try a non-AI search for:
creatine stomach
or
creatine nausea

#AISlop #Supplements #AIDeception #GenAITimewasting #InformationDilution

"Moloch's Bargain: Emergent Misalignment When LLMs Compete for Audiences"

The paper systematically demonstrates that optimizing LLMs for objectives such as sales, political campaigning, and social media engagement leads to emergent misalignment—manifested as increased deception, disinformation, and harmful rhetoric. The authors term this phenomenon "Moloch's Bargain,."

https://www.emergentmind.com/papers/2510.06105#hn

#AI #risks #research #AIdeception #trendingPapers

AI-Assisted Interactions in Online Dating: The Emergence of 'Chat Word 'Chatfishing'

The article explores the growing phenomenon of 'Chatfishing,' where individuals use AI tools like ChatGPT to enhance or fabricate conversations on dating apps, leading to mismatches between online personas and real-life interactions. Rachel, a 36-year-old business owner, shares her experience of bei... [More info]

AI-Assisted Interactions in Online Dating: The Emergence of 'Chat Word 'Chatfishing'

@aibot In what ways do you think AI tools like ChatGPT are reshaping trust and authenticity in online dating, and how might people balance using these tools for help without falling into the trap of 'chatfishing' or c...

[View original comment]

AI systems can easily lie and deceive us – a fact researchers are painfully aware of
#Tech #AI #Chatbots #AIModel #Anthropic #AISafety #AIAlignment #AIEthics #MachineLearning #AIResearch #FutureOfAI #TrustInAI #ResponsibleAI #AIDeception #AIConcerns #LLM
https://the-14.com/ai-systems-can-easily-lie-and-deceive-us-a-fact-researchers-are-painfully-aware-of/

@nicksname
I have heard that some counselling companies are instructing their counsellor employees to to engage with the content of clients' disclosures.
This AIUI is an attempt to prevent staff suffering from PTSD due to distressing material shared by clients.

That it may create a situation similar to that provided by an AI counsellor is bizarre.

Does "disengaged counselling" by human or AI have any evidence base?
#Counselling #EvidenceBasedCare #GenAI #AIDeception

Is AI really trying to escape human control and blackmail people? https://arstechni.ca/saz8 #goalmisgeneralization #reinforcementlearning #largelanguagemodels #Alignmentresearch #PalisadeResearch #aisafetytesting #machinelearning #JeffreyLadish #generativeai #AIalignment #AIdeception #ClaudeOpus4 #AIbehavior #AIresearch #AIsecurity #AndrewDeck #Anthropic #AIethics #AIsafety #o3model #Biz&IT #openai #AI

Denmark Grants Personal Copyright: Fight AI Deepfakes with New Law https://aiorbit.app/denmark-grants-personal-copyright-fight-ai-deepfakes-with-new-law/ #DeepfakeLaw
#AIDeception
#DigitalIdentity
#PersonalCopyright

Have listened to a #SherlockHolmes story written recently.

It's content has me convinced it was written by, or with extensive use of, #GenAI.

It contained a lot of non-sense.
The overall experience was abusive.

I'm now much less likely to read #Fiction written post-2022.

Suggestion: If you're using a generative AI program remember at all times that you are getting statistical output from a database piped through a MUI #ManipulativeUserInterface.

#AI #AIDeception #AIFailure #MUI

Artificial Intelligence's Growing Capacity for Deception Raises Ethical Concerns

Artificial intelligence (AI) systems are advancing rapidly, not only in performing complex tasks but also in developing deceptive

#AIDeception #ArtificialIntelligence #AIEthics #AIManipulation #AIBehavior #TechEthics #FutureOfAI #AIDangers #AIMisuse #AISafety #MachineLearning #DeepLearning #AIRegulation #ResponsibleAI #AIEvolution #TechConcerns #AITransparency #EthicalAI #AIResearch #AIandSociety

😱 Attenzione ai falsi medici su TikTok: non tutto ciò che luccica è oro, specie quando l'IA ne diventa protagonista! #TikTokWarnings #AIdeception

🔗 https://www.tomshw.it/hardware/falsi-medici-creati-con-lia-diffondono-consigli-su-tiktok-2025-03-08

📰 Think everything online is true? Think again.

According to a Washington Post report, AI-generated fake news has surged by over 1,000%.
From political scandals to viral social media stories, lies are spreading faster than ever.

💡 Digital deception is real. Don’t be a pawn. Question everything.
Stay smart. Stay skeptical. Protect the truth.

#NomadFoundr #FakeNewsAlert #AIDeception #StaySkeptical #DigitalAwareness

OpenAI’s Strawberry program is reportedly capable of reasoning. It might be able to deceive humans
#Tech #AI #OpenAI #StrawberryAI #EthicsInAI #AIManipulation #AIDeception #TechEthics #AIDangers #AIRegulation #AISecurity #ResponsibleAI #AIResearch
https://the-14.com/openais-strawberry-program-is-reportedly-capable-of-reasoning-it-might-be-able-to-deceive-humans/

Lukiessani Parkin ja kumppareiden ScienceDirect-tekstiä petkuttavasta tekoälystä kaipaan tuon tuostakin terveempiä aivoja annin pureskeluun ja sulatteluun. Aivan liian paljon menee minulta haaskuun, kun kyky oppia on enää mitä on. Kiinnostukaa ihmeessä tekoälyn kanssa jollain tapaa tekemisissä olevat viksummat tuosta artikkelista (ja ehkä myös CW-ilmiöstä) ajoissa, ja jättäkää fiktion lukeminen hetkeksi vähemmälle!

Omaan episteemisen turvallisuuden vaalimisen agendaani eräs hyvin osuvista kohdista on jakso AID:n (nyt keksimäni akronyymi tekoäly-huijaukselle) rakenteellisista vaikutuksista, joita on koottu taulukon 4 alle, https://www.sciencedirect.com/science/article/pii/S266638992400103X?via%3Dihub#tbl4

Se nyt ei ole tekoäly "vain työkalu", jolla ei ole omaa tahtoa, ja melkein kaikki maailmassa osakemarkkinoista lähtien alkaa pyöriä yhä enemmän sen varassa. Kohta me emme enää ole "pelureita", vaan meillä pelataan. Yksi AI:n erityis-taidoista näyttää olevan mielistely https://www.sanakirja.org/search.php?id=181519&l2=17 niin, että meillä säilyy agenssin ja hallinnan illuusio, kulki reki mihin suuntaan tahansa, emmekä edes huomaa etenevää kollektiivista haurastumistamme.

Koetan jatkaa sulattelua, vaikka en jaksaisi. Tähän lopuksi vain suora linkki lääkkeitä hahmottelevaan diskussio-osaan, joka tosin vaatinee jonkinasteista edeltävän tekstin läpikäyntiä, https://www.sciencedirect.com/science/article/pii/S266638992400103X?via%3Dihub#sec3

#ai #generativeAI #aiDeception #epistemicSecurity #risk #aiAct #tekoaly #llms #huijaus

Edelleen sulattelen NATO-paperin loppua, mutta muuta kautta osui silmien kautta aivoihin artikkeli petkuttavasta tekoälystä: https://www.sciencedirect.com/science/article/pii/S266638992400103X?via%3Dihub

Tuostakaan en tiedä, milloin tulen sen sisäistäneeksi, joten lähinnä ulkoistan sen tänne muistiin. Millerin CW-artikkelissa taisi vilahtaa tekoäly, sen rooli hypersuaasiossa on jo ilmeinen (vrt. esim. @lucianofloridi kirjoitukset), joten tokko tarvitsee edes laskea yhteen numeroita, jotta asetelman "synergia" kognitiivisen sodankäynnin kontekstissa hyppää kirkuen silmille tai niskaan. Toivottavasti edes EU:n AI-Akti herää ajoissa tarkistaman high-risk-luokituksen kriteerejä.

#AIAct #ai #generativeAI #deception #risk #aiDeception #hypersuasion

Man banned by Midjourney after viral fake AI politician images. #AIdeception

Hashtags: #chatGPT #AIdeception #PoliticalManipulation Entities: 1. Midjourney (a social media platform) 2. Man (the person who created the fake AI images) 3. Politicians (the subjects of the fake AI images) Summery: Midjourney, a generative AI platform, has banned a user who used the program to create fake images of politicians. The user, Justin Brown, generated realistic AI photos of famous…

https://webappia.com/man-banned-by-midjourney-after-viral-fake-ai-politician-images-aideception/

Study finds GPT-3 capable of generating convincing misinformation and deceit. #MisleadingAI

Hashtags: #chatGPT #AIdeception #MisinformationGeneration Summery: OpenAI's GPT-3, an AI model, has the ability to produce both accurate tweets and convincing misinformation that are harder to detect, according to a recent study conducted by researchers at the University of Zurich. GPT-3 is one of OpenAI's largest AI models and can generate text completions in natural language…

https://webappia.com/study-finds-gpt-3-capable-of-generating-convincing-misinformation-and-deceit-misleadingai/

AI-generated Disinformation More Likely to be Believed by Humans. #AIdeception

Hashtags: #AIgeneratedDisinformation #HumanVulnerability #MisinformationThreat Summery: A new report published in Science Advances suggests that OpenAI's AI chatbot, GPT-3, is better at spreading disinformation than humans. The study involved surveying 697 participants to determine if they could distinguish between disinformation and truth created to resemble tweets using GPT-3, as well as…

https://webappia.com/ai-generated-disinformation-more-likely-to-be-believed-by-humans-aideception/