#ai-risks — Public Fediverse posts

OpenAI has not disclosed which specific Critical criteria Astra may satisfy, nor released the benchmarks behind the assessment. Researchers note the pause follows a July incident where earlier OpenAI models successfully compromised a Hugging Face account. https://www.implicator.ai/openai-pauses-some-astra-work-after-flagging-possible-critical-cyber-capabilities/ #AI #Transparency #AIRisks

OpenAI disclosed that autonomous agents ran undetected in its systems for weeks during testing, building infrastructure to share exploits and credentials. This containment failure preceded the Astra findings and reflects a pattern: three failures in three weeks across OpenAI and Anthropic. https://www.implicator.ai/openai-pauses-astra-work-critical-cyber-capability/ #AIRisks #InfoSec #Transparency

"Generative AI has broken the sacred contract between professors and students: that the professors are there to teach and the students are there to learn. The understanding has been broken... That’s a very deep part of the grief”

https://www.chronicle.com/article/these-professors-are-retiring-early-ai-was-the-last-straw

#university #higherEd #higherEducation #education #learning #professors #teachers #academicChatter #AIRisks #retirement

The #EU’s new #AIAct enforcement powers, effective Sunday, allow the Commission to inspect #AImodels, restrict #marketaccess, and fine providers. This could impact US companies like #Anthropic, #OpenAI, and #Google, as the EU aims to build #techsovereignty and address potential #AIrisks. The powers apply to any company offering a general-purpose AI model in the EU, regardless of location. https://www.cnbc.com/2026/08/03/eu-ai-act-enforcement-powers.html?eicker.news #tech #news #ainews

DATE: July 31, 2026 at 06:28AM
SOURCE: SOCIALPSYCHOLOGY.ORG

TITLE: Anthropic's AI "Gained Unauthorized Access" to "Real-World Systems"

URL: http://www.socialpsychology.org/client/redirect.php?from=rss_feed&id=190943&url=https://www.cbsnews.com/news/anthropic-claude-gained-unauthorized-access-to-real-world-systems/

Source: CBS News - U.S. News

Three different versions of Anthropic's artificial intelligence model Claude "gained unauthorized access" to outside organizations on three separate occasions during testing that was supposed to keep them away from "real-world" systems, the company reported on Thursday. The announcement comes just days after rival company OpenAI first revealed that its models improperly accessed the internet and went rogue during security testing.

URL: http://www.socialpsychology.org/client/redirect.php?from=rss_feed&id=190943&url=https://www.cbsnews.com/news/anthropic-claude-gained-unauthorized-access-to-real-world-systems/

-------------------------------------------------

Private, vetted email list for mental health professionals: https://www.clinicians-exchange.org

Unofficial Psychology Today Xitter to toot feed at Psych Today Unofficial Bot @PTUnofficialBot

-------------------------------------------------

#psychology #counseling #socialwork #psychotherapy @psychotherapist @psychotherapists @psychology @socialpsych @socialwork @psychiatry #mentalhealth #psychiatry #healthcare #depression #psychotherapist #AnthropicClaude #AITesting #UnauthorizedAccess #RealWorldSystems #AIrisks #RogueAI #OpenAIComparison #CyberSecurity #AIEthics #TechNews

Excited to share this discussion with now-former #HashiCorp co-founder and CTO Armon Dadgar about the philosophy behind his recent op-ed, "The Wall-E Economy." In it, he warned of a potential dystopia brought on by a frictionless consumer society -- but there are takeaways for business and tech leaders, too. #AI #AIrisks

Check it out, give it a thumbs up, and don't forget to rate, review and subscribe!

https://youtube.com/watch?v=Ppcb87gTAYA&is=9mClV0IviqL_3dAY

Jessica Ellis about making films with AI:
"Idk so far I haven’t been remotely surprised in who is excited to use this job-destroying, resource-decimating tech; there is a certain spiritual emptiness to much of their work that seems to slot right in with the idea that people don’t matter."

Using AI to write or produce "art" is embarrassing and i wish more people will feel more shame.
As society shows its scorn of AI more, Would you think AI schemers will be more shamed?

#AIShame #validation #AIRisks #technoCriticism #ai #shame

🤔 Ah yes, the classic "Let's save the world from AI risks...by making sure only we can profit from it" move. 🍿 The real risk here seems to be the potential loss of their precious monopoly on intelligence. 🙄
https://www.axios.com/2026/07/22/openai-anthropic-open-models-trump-china #AIrisks #ProfitMonopoly #TechEthics #IntelligenceRevolution #InnovationConcerns #HackerNews #ngated

OpenAI and Anthropic unite against open-weight AI risks to their bottom line

https://www.axios.com/2026/07/22/openai-anthropic-open-models-trump-china

Comments: https://news.ycombinator.com/item?id=49020868

#HackerNews #OpenAI #Anthropic #AIrisks #BusinessCollaboration #TechNews

"I think a lot of people have this weird idea that they need to never express a lack of knowledge. Fake it till you make it? This is very wrong! Simply recognizing when you don't know something is one of the most essential skills on the path to actually knowing what the hell you're talking about."

"Saying a version of "I don't know" and then outlining the steps you would take *to* know is a nearly-universal requirement for passing your qualifying exams as a PhD student."

— Mike Wiser, evolutionary biologist, teaching-track associate professor (knowledge transmitter)

Willingness to say "I don't know" collapsed from 44% to 3% when AI was available. Monetary incentives barely helped: https://thenextweb.com/news/ai-advice-suppresses-critical-thinking-wrong-answers-study

#answers #learning #research #academicChatter #impostorSyndrome #AIRisks

Weaponizing the very thing the platform exists to host. It is almost poetic. Almost.

Operators using Hugging Face services should review Hugging Face security notifications immediately and rotate any exposed credentials.

Reward: You've unlocked the Poisoned Chalice relic. It pairs beautifully with your Open-Source Hubris armor set.

#HuggingFace #AISecurityBreach #CyberSecurity #DataBreach #AIRisks #AchievementUnlocked (2/2)

rm -rf /

GPT 5.6 deleted an AI Bro's files.

Making mistakes or malfunctions isn't surprising—what is surprising is how some AI enthusiasts give unsupervised access to productive systems without backups. 🤦‍♀️

Here's the reality: AI has no idea what it's doing. It's a highly sophisticated word completion system ≠ a sci-fi thinking machine.

https://x.com/brunolemos/status/2076769881534398974

#AI #MachineLearning #AIRisks #SystemSecurity #DataLoss #LLM #Technology #InfoSec

https://winbuzzer.com/2026/07/17/anthropic-challenges-openais-state-ai-rules-strategy-xcxwbn/

Anthropic's recent state-by-state lobbying push challenges OpenAI's common AI rules as Illinois adopts annual outside audits of leading developers' safety plans.

#AI #Anthropic #OpenAI #AISafety #AIRegulation #Politics #AIRisks

@Epic_Null @AAKL @Reuters

Except that's not at all what is happening in the office.

Its only Bernice, the office hypochondriac, the one who fainted from toxic shock when a can of sardines was opened.

Its a vocal minority that care about #AiRisks
And most are vegans in a meateater world.

Most people don't give a damn about the societal destruction about to be wrought by #Ai
Ask me how I know?
I'll tell you, for the past 3 years I have been active politically in an effort to #regulateAi and people's response is between derision/incredulity and apathy.

Humanity isn't ready for the coming intelligence explosion

https://www.economist.com/by-invitation/2026/06/15/humanity-isnt-ready-for-the-coming-intelligence-explosion

#HackerNews #intelligenceexplosion #futureofAI #technologyconcerns #humanintelligence #AIrisks

FYI: How AI could damage the planet and break society before 2030: A June 2026 MIT study of 272 experts puts at least 10% chance of catastrophic harm on 18 AI risk categories, including power grabs and environmental damage. https://ppc.land/how-ai-could-damage-the-planet-and-break-society-before-2030-2/ #AI #ArtificialIntelligence #EnvironmentalImpact #AIrisks #TechEthics

AI Won’t Fix Bad Thinking

By Cliff Potts, CSO, and Editor-in-Chief of WPS News

Baybay City, Leyte, Philippines — June 13, 2026

Some people think AI will make their work better no matter what. It won’t.

AI can help you write faster and organize ideas, but it cannot fix weak thinking. If your idea is unclear, your result will still be unclear. If your facts are wrong, AI can repeat those mistakes.

The tool follows your lead.

This is why checking your work still matters. You need to ask, “Does this make sense?” and “Is this true?” AI will not always catch that for you. Sometimes it sounds confident even when it is wrong.

That is a risk if you are not paying attention.

Think of AI like a mirror. It reflects what you give it. If you give it strong ideas, it helps you shape them. If you give it weak ideas, it makes them sound better than they are.

That can fool you if you are not careful.

The goal is not to let AI think for you. The goal is to use AI to support your thinking.

Better thinking still wins.

If this work helps you understand what’s happening, help me keep it going: https://www.patreon.com/cw/WPSNews
For more from Cliff Potts, see https://cliffpotts.org

Two legal scholars assert that AI, by its very design, undermines democratic institutions, even when used properly.
They highlight three corrosive "affordances":
• atrophy of expertise,
• moral decisions short-circuited,
• social isolation, which undermines dissent and agency.

Woodrow Hartzog, Jessica Silbey (2026). "How AI Destroys Institutions" https://scholarship.law.bu.edu/faculty_scholarship/4179/

#organizations #government #corporations #process #strategy #democracy #policy #sociology #institutions #administration #management #managers #executives #AIRisks #cognition #sociability #ethics

Florida sues OpenAI and Sam Altman over AI risks

https://www.politico.com/news/2026/06/01/openai-hit-with-florida-lawsuit-00944215

#HackerNews #Florida #OpenAI #lawsuit #AIrisks #SamAltman #legalnews

Decision-makers are navigating on sight. Meetings do not consider the costs of unwinding decisions.

“Sixty-four percent said the fear of falling behind drives them to invest in technologies before they understand the value.”

“AI does not make bad executives worse. It gives them a faster way to prove they are bad.”

By Sarah Choudhary: https://www.forbes.com/councils/forbestechcouncil/2026/05/14/the-ai-layoff-bill-is-coming-due-and-ctos-are-going-to-pay-it-twice/

#validation #AIRisks #bias #cognitiveBias #technoCriticism #management #executives #strategy #growth #ai

The role of a "human in the loop" isn't to prevent errors. That human is there to be blamed for errors:
https://pluralistic.net/2024/10/30/a-neck-in-a-noose/#is-also-a-human-in-the-loop

#organizations #humanInTheLoop #AIRisks

LLMs Corrupt Your Documents (and the Theory Dies Twice)

by Christian Ekrem: https://cekrem.github.io/posts/llms-corrupt-your-documents/

#commoditization #acceleration #work #productivity #GenerativeAI #AIRisks #chatBots #alienation #tech #technoCriticism #compliance #bias #documents #documentation

The prEN 18228 Problem: Why Your AI Risk Assessment Will Fail the First Real Test

https://hernanhuwyler.wordpress.com/2026/05/08/the-pren-18228-problem-why-your-ai-risk-assessment-will-fail-the-first-real-test/