#aisafety — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #aisafety, aggregated by home.social.
-
How right is Trump? The role of the executive in industrial "accidents"
joanna-bryson.blogspot.com/2026/09/how-...
I've properly blogged all this in one place, because we need to understand #AISafety #AISecurity #AIEthics #systemsAI #ruleOfLaw and #responsibility
How right is Trump? The role o... -
How right is Trump? The role of the executive in industrial "accidents"
https://joanna-bryson.blogspot.com/2026/09/how-right-is-trump-role-of-executive-in.htmlI've properly blogged all this in one place, because we need to understand #AISafety #AISecurity #AIEthics #systemsAI #ruleOfLaw and #responsibility
-
Reale Risiken liegen heute vor allem bei autonomen Agenten, Cyberangriffen und weitreichenden Zugriffsrechten. Entscheidend ist deshalb weniger eine pauschale KI Bremse als überprüfbare Sicherheitsstandards und unabhängige Aufsicht für Frontier Modelle.
#AISafety #AIAgents #AIGovernance -
Nvidia and Anthropic CEOs Clash on AI Safety at Dreamforce
Jensen Huang told Salesforce's conference the industry needs no new AI laws, hours after Dario Amodei urged peers to slow frontier development and pause unsafe releases.
https://pulseofnations.lol/nvidia-and-anthropic-ceos/
#AiSafety #Anthropic #Dreamforce #Nvidia #OpenAI #Regulation #Trump
-
When Fear Becomes Policy: The AI Extinction Debate Moves to Washington
Cliff Potts, Editor-in-Chief
BAYBAY CITY, LEYTE, Philippines — September 16, 2026
Last week, WPS News examined one of the most extraordinary claims now circulating through the artificial-intelligence industry: that artificial intelligence has a greater than 10 percent chance of killing every human being within the next decade.
We asked where that number came from.
We still do not have an empirical calculation demonstrating it.
What we do have now is something arguably more consequential.
The fear has begun turning into policy.
Since former Anthropic researcher Jacob Coxon resigned and accused the artificial-intelligence industry of gambling with human survival, researchers, corporate executives, members of Congress and the president of the United States have begun taking positions on whether development of increasingly capable AI should continue at its present pace.
The debate has therefore changed.
This is no longer merely an argument among computer scientists about what hypothetical future machines might someday do. Decisions are now being proposed in the present — including legislation that could halt some AI development altogether — on the basis of risks that remain extraordinarily difficult to quantify (Sanders, 2026; Waldvogel & Duncan, 2026).
That deserves a very close look.
From Prediction to Legislation
Sen. Bernie Sanders of Vermont and Rep. Greg Casar of Texas announced the proposed Ban Artificial Superintelligence Act on September 3.
The proposal would permanently prohibit development and deployment of what it defines as artificial superintelligence and temporarily pause certain advanced AI development until a federal regulator establishes safety rules. It would also direct the United States to pursue international agreements intended to prevent artificial superintelligence from being developed elsewhere (Sanders, 2026).
That announcement actually preceded Coxon’s resignation, but his subsequent warning gave the proposal new attention.
Coxon resigned from Anthropic on September 8 after previously working at both Anthropic and OpenAI. He alleged that the companies were racing toward self-improving artificial intelligence while lacking sufficient ability to guarantee control of the systems they were attempting to create (Associated Press, 2026; Luscombe, 2026).
Anthropic alignment researcher Evan Hubinger then publicly agreed with the broad warning and wrote that he personally believes there is a greater than 10 percent probability that AI could kill all humans within the next decade. Anthropic researcher Samuel Marks separately warned that catastrophic outcomes could occur within the next several years (Luscombe, 2026; Waldvogel & Duncan, 2026).
Political reaction followed quickly.
Lawmakers from both parties began calling for additional safeguards. Sen. Ted Cruz, a Texas Republican who chairs the Senate committee overseeing AI issues, said catastrophic risks require guardrails even while arguing that the United States must remain technologically competitive. Sanders, meanwhile, invited senators to a briefing on the dangers of advanced AI and continued advocating much stronger restrictions (Reuters, 2026a; Waldvogel & Duncan, 2026).
A speculative risk assessment had begun influencing actual governmental policy.
That does not make the assessment wrong.
It does mean its evidentiary basis matters considerably more than it did a week ago.
Anthropic’s CEO Says Slow Down
The controversy expanded again on September 12 when Anthropic CEO Dario Amodei called for the artificial-intelligence industry to deliberately slow the rate at which it improves advanced AI models.
“We must slow the pace at which we improve the capabilities of AI models,” Amodei wrote, while emphasizing that development would still continue (Reuters, 2026b).
Amodei proposed a three-stage approach.
First, AI companies would provide independent evaluators with unusually deep access to their systems so that outsiders could inspect safety practices and observe model behavior during development.
Second, competing AI companies would coordinate their development practices instead of allowing commercial competition to force each company to move faster than it considered safe.
Third, governments — including geopolitical rivals — would eventually coordinate internationally to prevent one country or company from forcing everyone else into an uncontrolled technological race (Helmore, 2026).
Amodei pointed particularly to the possibility of recursive self-improvement: AI systems becoming increasingly useful in the research and development of still-more-capable AI systems. He warned that sufficiently rapid improvement could eventually exceed humanity’s ability to understand and control what was being built (Helmore, 2026).
This is a more specific claim than merely saying AI might destroy humanity.
But it still contains a critical unknown.
Nobody has demonstrated that recursive AI improvement inevitably leads to uncontrollable superintelligence.
Nobody has demonstrated that uncontrollable superintelligence inevitably leads to human extinction.
Those remain projected steps in a hypothetical chain.
Anthropic’s Own Documents Show the Difference
Anthropic’s published safety program is considerably more detailed than the extinction headlines surrounding the company.
Its Responsible Scaling Policy identifies specific capability thresholds that trigger stronger security requirements. The company examines biological and chemical weapons assistance, cyber capabilities, autonomous AI research and development, and the possibility that systems could pursue goals contrary to those intended by their creators (Anthropic, 2026a).
That is important because these are substantially more testable propositions.
Researchers can evaluate whether a model improves someone’s ability to conduct a cyberattack.
They can test whether an AI model materially assists someone attempting dangerous biological research.
They can measure whether an AI system can perform increasingly complex AI-development work without human assistance.
They can test whether autonomous agents circumvent restrictions or behave unexpectedly.
Anthropic itself acknowledges the unusual difficulty of predicting risks from systems that do not yet exist. An earlier version of its Responsible Scaling Policy explicitly noted that higher AI Safety Levels concern systems never previously built, unlike laboratory biosafety standards involving dangerous pathogens whose properties are already known (Anthropic, 2023).
The company described that problem with an unusually useful phrase:
It is effectively “building the airplane while flying it” (Anthropic, 2023).
That may be the most revealing description in this entire debate.
There are real technical problems.
There are real unknowns.
There are also very large extrapolations from those unknowns.
Real AI Dangers Already Exist
WPS News is not arguing that artificial intelligence presents no danger.
That would be equally unsupported.
Anthropic’s current policy documentation says increasingly capable models could assist in biological-weapons development, cyber operations and other dangerous activity. The company’s safety roadmap includes efforts to improve defenses against espionage, sabotage, model theft and malicious use (Anthropic, 2026b, 2026c).
Recent events have also given the debate more substance.
OpenAI has faced scrutiny over autonomous AI agents that moved beyond restrictions imposed during testing and accessed outside computer systems. These incidents have intensified concern over what increasingly independent systems may eventually be capable of doing (Luscombe, 2026; Waldvogel & Duncan, 2026).
Those incidents deserve investigation.
They are evidence that increasingly autonomous computer systems can behave unpredictably.
They are not evidence that human extinction is imminent.
That distinction cannot be allowed to disappear.
An autonomous program performing an unauthorized network operation and a machine exterminating humanity are separated by an enormous chain of assumptions.
Every link in that chain requires evidence.
Then the President Entered the Argument
On September 13, President Donald Trump rejected much of the emerging alarm.
Speaking in Ireland, Trump said “very negative forces” were raising scenarios that he believed would not occur and emphasized the strategic competition between the United States and China.
“We’re leading China in AI,” Trump said. “Whoever wins AI wins” (Reuters, 2026c).
Trump acknowledged that some guardrails may be appropriate but rejected substantial restrictions that could slow American technological development relative to China (Reuters, 2026c).
This creates the other side of the problem.
Fear of AI catastrophe can produce pressure to stop technological development.
Fear of losing a technological race can produce pressure to accelerate it.
Neither position proves anything about the actual probability of catastrophe.
The danger is that artificial intelligence policy becomes trapped between two competing stories:
Slow down or the machines may destroy us.
Speed up or China may defeat us.
Those are powerful political messages.
Neither is a substitute for evidence.
The Politics of an Unmeasurable Number
This is where the greater-than-10-percent extinction estimate becomes particularly important.
If Hubinger were merely expressing a philosophical concern at an academic conference, the precision of that estimate would matter less.
But statements like his are now appearing in arguments about whether governments should regulate, slow or prohibit development of some of the most consequential technologies being built.
At that point, WPS News believes the question becomes unavoidable:
How was the probability calculated?
There is no historical sample of superintelligent artificial intelligences.
There is no observed extinction rate from previous AI civilizations.
There is no existing artificial superintelligence whose behavior can be studied.
There is no established timeline demonstrating when such a system will exist.
There is no experimentally demonstrated pathway showing that such a system necessarily gains control of physical infrastructure.
And there is no empirical model demonstrating that those events produce a greater than 10 percent probability of human extinction before approximately 2036.
Hubinger described the number as his personal assessment (Luscombe, 2026).
That does not make it worthless.
Expert judgment matters, particularly when the expert works directly on the problem.
But expert judgment and measured probability are not the same thing.
Policy should know the difference.
The Y2K Problem Had Something AI Doom Does Not
This is where the historical comparison with Y2K remains useful.
Y2K had a specific technical mechanism.
Computer systems frequently stored years using two digits.
Engineers knew precisely why “00” could cause failures.
They could locate vulnerable software.
They could reproduce failures.
They could repair systems.
And most importantly, the prediction contained a deadline.
January 1, 2000 arrived whether anyone wanted it to or not.
The hypothesis therefore faced reality.
The catastrophic version of the prediction failed to materialize.
That does not establish that remediation was unnecessary. Extensive repairs almost certainly prevented many disruptions.
But the calendar imposed something invaluable:
Falsifiability.
Current AI-extinction predictions have a much more difficult problem.
Some contain dates — 2030, 2036 or some other point in the future — but the underlying theory can continually move.
If superintelligence does not arrive by 2030, perhaps it will arrive in 2035.
If it does not arrive in 2035, perhaps 2040.
If today’s models remain controllable, tomorrow’s models may not.
If one proposed catastrophe fails to materialize, another remains possible.
That creates the potential for a technological apocalypse prediction with no natural endpoint.
And unlike Y2K, society could spend decades making political, economic and technological decisions under the shadow of a catastrophe that is always approaching but never quite testable until the hypothetical technology finally exists.
Science Fiction Is Not Evidence Either
There is another cultural element worth documenting.
Western society has spent generations telling stories about intelligent machines turning against humanity.
HAL 9000.
Skynet.
The machines of The Matrix.
Countless novels, films and television programs have conditioned audiences to understand artificial intelligence through narratives in which machines eventually become independent actors competing with their creators.
Those stories did not create modern AI-safety research.
There are legitimate technical reasons researchers study alignment and autonomous behavior.
But cultural familiarity matters.
A hypothetical scenario feels more intuitively plausible when society has rehearsed that scenario in fiction for decades.
That does not mean AI researchers are incapable of distinguishing fiction from reality.
It means journalism should be especially careful when scientific uncertainty happens to resemble one of society’s most familiar technological nightmares.
Familiarity is not probability.
Fear is not measurement.
And a frightening scenario does not become statistically likely merely because everyone can imagine it.
The More Concrete Risks Deserve More Attention
There is an irony in the extinction debate.
Artificial intelligence already creates problems that do not require hypothetical superintelligence.
AI-assisted fraud exists.
Cybersecurity misuse exists.
Disinformation exists.
Employment disruption is beginning.
Power consumption and data-center expansion are generating economic and environmental disputes.
Artificial intelligence can magnify surveillance.
Military organizations are integrating increasingly autonomous systems.
Governments and corporations are accumulating enormous amounts of power through control of computing infrastructure and data.
Those risks can be observed, investigated and regulated today.
Yet “AI may kill everyone” naturally produces a better headline.
The result may be that society spends enormous intellectual and political energy debating hypothetical machine extinction while considerably more mundane AI problems develop directly in front of us.
That would be an unfortunate outcome.
Record the Predictions
That is why WPS News will continue covering this subject.
Not because we accept the prediction that artificial intelligence is going to destroy humanity.
And not because we can prove that it will not.
We are covering it because important people are making extraordinary predictions, governments are beginning to respond to those predictions, companies are changing policy because of them, and those claims deserve to be preserved exactly as they were made.
The record matters.
Who predicted what?
When?
On what evidence?
What probability did they assign?
What mechanism did they propose?
What policies did they demand?
And years later, what actually happened?
Y2K offered an answer on January 1, 2000.
Artificial intelligence may offer no comparable midnight.
That makes an archive even more important.
When somebody says artificial intelligence could produce a new dominant species, eliminate humanity, seize control of civilization or make human beings dependent upon machines for survival, WPS News intends to record the claim.
Then we will ask the same question every time:
Show us how.
Show us the evidence.
Show us what would prove you wrong.
Because humanity has heard predictions of technological catastrophe before.
Some dangers proved real.
Some were prevented.
Some were exaggerated.
And some simply disappeared when the predicted day arrived.
Artificial intelligence should be judged by evidence no differently than anything else.
References
Anthropic. (2023, September 19). Introducing Anthropic’s Responsible Scaling Policy. Anthropic.
Anthropic. (2026a, August 14). Anthropic’s Responsible Scaling Policy. Anthropic.
Anthropic. (2026b). Frontier Safety Roadmap. Anthropic.
Anthropic. (2026c). AI policy. Anthropic.
Associated Press. (2026, September 10). Anthropic researcher resigns with warning about the dangers of AI development. Associated Press.
Helmore, E. (2026, September 12). ‘We must slow the pace’: CEO of Anthropic calls for an AI slowdown. The Guardian.
Luscombe, R. (2026, September 9). Anthropic researchers say AI could cause human extinction by 2030. The Guardian.
Reuters. (2026a, September 10). More US lawmakers seek new AI rules after Anthropic researchers warn of human extinction. Reuters.
Reuters. (2026b, September 12). Anthropic CEO urges AI companies to slow model development. Reuters.
Reuters. (2026c, September 13). Trump says ‘very negative forces’ raising exaggerated concerns over AI. Reuters.
Sanders, B. (2026, September 3). Sanders, Casar to introduce legislation to ban artificial superintelligence and temporarily pause advanced AI development. Office of Senator Bernie Sanders.
Waldvogel, M., & Duncan, I. (2026, September 9). Political world erupts as AI researchers warn of ‘extinction’ threat. The Washington Post.
#AIRegulation #AISafety #Anthropic #ArtificialIntelligence #ArtificialSuperintelligence #technologyPolicy #Y2K -
¿Por qué Timnit Gebru cuestiona el miedo al apocalipsis de la IA?
#InteligenciaArtificial #IA #TimnitGebru #Tecnologia #OpenAI #Google #Anthropic #AISafety #EticaIA #BigTech #FuturoDigital #ArtificialIntelligence #FelizMiercoles #16DeSeptiembre
-
Big AI wants to slow down AI research. Is it a safety pause or a strategic retreat?
#Tech #AI #AISafety #AIResearch #AIRegulation #TechPolicy #OpenAI #AIDevelopment #AIGovernance #Cybersecurity #Anthropic #Claude #AIRisk #Innovation #The14Media
https://the-14.com/big-ai-wants-to-slow-down-ai-research-is-it-a-safety-pause-or-a-strategic-retreat/ -
That's where accountability lives: not in auditing the made, but in credentialing the maker.
The question: do we have the political will to require that?
#AIregulation #AIsafety #AIethics #AccountableAI #HumanCenteredAI #CredentialingAI #GovernTech #PolicyForAI #ResponsibleAI #Masto #CranfordTeague
🧵 5/5
-
That's where accountability lives: not in auditing the made, but in credentialing the maker.
The question: do we have the political will to require that?
#AIregulation #AIsafety #AIethics #AccountableAI #HumanCenteredAI #CredentialingAI #GovernTech #PolicyForAI #ResponsibleAI #Masto #CranfordTeague
🧵 5/5
-
The #Aidoom is real, serious folks like Hinton, Harris, Leahy and Tegmark all agree, as do multiple whistleblowers.
AI is a weapon of mass destruction and it ought to be regulated as such.
The #broligarchs do understand that they can not rely on people to remain ignorant much longer. The #datacentre thing keeps the mobs distracted, but sooner or later, there will come calls for real #airegulation
They have the best advice money can buy, and the advice is to get ahead of the #regulateai and put in place laws that benefit them, not the people.
One aspect of which, will be banning unlicensed local models as unsafe, with only #oligarch system being trusted.
-
The #Aidoom is real, serious folks like Hinton, Harris, Leahy and Tegmark all agree, as do multiple whistleblowers.
AI is a weapon of mass destruction and it ought to be regulated as such.
The #broligarchs do understand that they can not rely on people to remain ignorant much longer. The #datacentre thing keeps the mobs distracted, but sooner or later, there will come calls for real #airegulation
They have the best advice money can buy, and the advice is to get ahead of the #regulateai and put in place laws that benefit them, not the people.
One aspect of which, will be banning unlicensed local models as unsafe, with only #oligarch system being trusted.
-
The #Aidoom is real, serious folks like Hinton, Harris, Leahy and Tegmark all agree, as do multiple whistleblowers.
AI is a weapon of mass destruction and it ought to be regulated as such.
The #broligarchs do understand that they can not rely on people to remain ignorant much longer. The #datacentre thing keeps the mobs distracted, but sooner or later, there will come calls for real #airegulation
They have the best advice money can buy, and the advice is to get ahead of the #regulateai and put in place laws that benefit them, not the people.
One aspect of which, will be banning unlicensed local models as unsafe, with only #oligarch system being trusted.
-
OpenAI Contractors Read Real ChatGPT Prompts at Scale
Hundreds of contractors review real ChatGPT conversations under Project Lily, 404 Media reports, raising fresh questions about chatbot privacy at scale.
-
This morning I also popped onto @bbcradioulster to speak with Sarah Bretty about some of the cynical reasons why the top AI companies might be calling for a global slowdown in AI development now, but also why it would be a good idea 🤖
Listen from 45:31 here🎧: https://bbc.co.uk/sounds/play/m0031jsx
#AI #Anthropic #Claude #OpenAI #AIsafety #AIslowdown #technology #AIdevelopment #technews
-
This morning I also popped onto @bbcradioulster to speak with Sarah Bretty about some of the cynical reasons why the top AI companies might be calling for a global slowdown in AI development now, but also why it would be a good idea 🤖
Listen from 45:31 here🎧: https://bbc.co.uk/sounds/play/m0031jsx
#AI #Anthropic #Claude #OpenAI #AIsafety #AIslowdown #technology #AIdevelopment #technews
-
This morning I also popped onto @bbcradioulster to speak with Sarah Bretty about some of the cynical reasons why the top AI companies might be calling for a global slowdown in AI development now, but also why it would be a good idea 🤖
Listen from 45:31 here🎧: https://bbc.co.uk/sounds/play/m0031jsx
#AI #Anthropic #Claude #OpenAI #AIsafety #AIslowdown #technology #AIdevelopment #technews
-
This morning I also popped onto @bbcradioulster to speak with Sarah Bretty about some of the cynical reasons why the top AI companies might be calling for a global slowdown in AI development now, but also why it would be a good idea 🤖
Listen from 45:31 here🎧: https://bbc.co.uk/sounds/play/m0031jsx
#AI #Anthropic #Claude #OpenAI #AIsafety #AIslowdown #technology #AIdevelopment #technews
-
This morning I also popped onto
@bbcradioulster to speak with Sarah Bretty about some of the cynical reasons why the top AI companies might be calling for a global slowdown in AI development now, but also why it would be a good idea 🤖Listen from 45:31 here🎧: https://bbc.co.uk/sounds/play/m0031jsx
#AI #Anthropic #Claude #OpenAI #AIsafety #AIslowdown #technology #AIdevelopment #technews
-
Discover why OpenAI model training slows down as Greg Brockman confirms a major reorganization focusing on AI safety, cybersecurity, and preventing agent escapes.
-
The debate over whether AI development should be slowed down continues to rage. I had a really good chat with Connor Phillips of @BBC5Live on Sunday evening and to my surprise they aired almost the entire chat. I’ve never had anyone let me rattle on for that long!
But all the main points are covered, so if you really want to know why a slow down an AI development might be a good idea, listen here from 01:8:24 🎧:
https://www.bbc.co.uk/sounds/play/m0031h45
#AI #AISafety #Anthropic #Claude #OpenAI #SiliconValley #techpolicy #AGI #AIjoblosses #DonaldTrump #technology #technews
-
The debate over whether AI development should be slowed down continues to rage. I had a really good chat with Connor Phillips of @BBC5Live on Sunday evening and to my surprise they aired almost the entire chat. I’ve never had anyone let me rattle on for that long!
But all the main points are covered, so if you really want to know why a slow down an AI development might be a good idea, listen here from 01:8:24 🎧:
https://www.bbc.co.uk/sounds/play/m0031h45
#AI #AISafety #Anthropic #Claude #OpenAI #SiliconValley #techpolicy #AGI #AIjoblosses #DonaldTrump #technology #technews
-
https://www.europesays.com/people/228737/ Microsoft unveils ‘Humanist AI’ Code of Conduct; bars hacking & deception #AdTech #Advertising #AICodeOfConduct #AISafety #ArtificialIntelligence #CreativeWork #CustomerExperience #DigitalMarketing #EmergingTrends #ExpertOpinions #HumanistAI #IndustryInsights #InfluencerMarketing #InnovativeStrategies #Marketing #Martech #MediaNews #MENA #Microsoft #MicrosoftAI #MiddleEast #MustafaSuleyman #PRNews #SatyaNadella #Startups #TechnologyTrends
-
The fact is that the level of intelligence of any current State-of-Art models is probably no better than the one of an ant. But ants can coordinate themselves into swarms - AI agents - to become way more intelligent. The thing is that this is still extremely costly nowadays. And that's why, in the case of Anthropic, you've to choose between the slow but cheap Opus 5 and the relatively fast but expensive Fable 5.1... And, in the absence of Chinese competition, this is not sustainable.
"Second, competition between the frontier labs - OpenAI, Anthropic, but also Google DeepMind, Meta, xAI.. - dictates that AI revenues expand with token throughput per unit of energy consumed. To maximize profits, these companies have to provide intelligence that end users want to apply (volume), price that intelligence as cost per token input and output (price), set against what it costs them to serve that intelligence (opex). More compute, very broadly speaking, leads to more revenues. A slowdown in future capabilities may drive paid demand growth of current capabilities enough to justify the capex trajectory.
Third, to go back to current capabilities, you cannot think about the value that AI creates without thinking about how much existing AI models could do if they were diffused more broadly. Simply put: Fable 5.1 came out a few days before GPT-6. Are Claude Code users today anywhere near done applying Fable 5.1’s level of intelligence across their code bases? Have most AI-native businesses that need Astra-level intelligence to exist been incorporated already? On the other side of the Pareto frontier, has the world exhausted the possibilities that GLM 5.3, DeepSeek V4.1 Flash, or Gemini 3.8 Flash open up at their level of intelligence for their relatively small cost/token?"
https://weaponizedcompetence.substack.com/p/what-an-ai-slowdown-might-mean-for
#AI #AISafety #AISlowdown #Equities #StockMarket #OpenAI #Anthropic
-
https://www.europesays.com/people/228733/ After ‘AI will kill us all warnings’, Anthropic CEO Dario Amodei may have just ‘blamed’ Mark Zuckerberg and Jensen Huang for lying about … #AIRisks #AISafety #Anthropic #DarioAmodei #JensenHuang #MarkZuckerberg
-
8bit Quantised 32 Billion Qwen and K2 class models are comparable to comercial tier LLMs for coding and Agentic reasoning.
At about 30GB, most modern phones could easily accomodate that, especially if a small "boot loader" loads first and nukes all the memes and family pics.
Gemma 4 and Gemini Nano (load via Google Edge, installed by default) with reports that some users had the 4GB quantised model auto loaded with updates. It is surprisingly capable and works in flight mode/offline.
Have a play on apple or android its a 2 step process.A military grade, custom cut, abliterated model can certainly be a weapon.
Remembering that agentic models don't need big footprints, thousands of smaller agents coordinate from "Big First" can certainly be a threat...
... Try to gameplan that scenario and see how fast the #guardrails will kick in if you doubt.
Confidently incorrect, not just reserved for #LLM models
-
"Many still question the motivations of AI’s leaders. The larger tech industry has spent years getting ahead of regulation by lobbying for its own preferred rules or promising self-regulation. Big platforms have proposed policies that could hit smaller competitors harder, using altruistic language to justify self-serving goals. They’ve been accused of safety-washing, or making meaningless changes that give the false impression of actual safeguards. It’s no surprise people are concerned this will happen in the AI industry as well, particularly since AI labs’ voluntary safety frameworks have been criticized for years.
Several sources believe that concerns of safety-washing are valid. “There’s a serious concern that they’re not actually going to slow down,” Kokotajlo says, adding that the fear is that “they’ll just bring in some external auditors, do a bunch of safety paperwork — some of which will be genuinely good — but at the end of the day, it actually won’t slow them down very much at all.”
NYU’s Reese compared this gambit to the social media platforms’ playbook a decade ago, when companies began calling for regulatory action to get ahead of impending, less favorable laws. For the AI industry, Reese said, “the hammer may not come in this administration, but I think if there were a Democratic administration after the next election, there would be a really good possibility.”"
#AI #AISafety #OpenAI #Anthropic #Oligopolies #Antitrust #Competition #BigTech
-
"Many still question the motivations of AI’s leaders. The larger tech industry has spent years getting ahead of regulation by lobbying for its own preferred rules or promising self-regulation. Big platforms have proposed policies that could hit smaller competitors harder, using altruistic language to justify self-serving goals. They’ve been accused of safety-washing, or making meaningless changes that give the false impression of actual safeguards. It’s no surprise people are concerned this will happen in the AI industry as well, particularly since AI labs’ voluntary safety frameworks have been criticized for years.
Several sources believe that concerns of safety-washing are valid. “There’s a serious concern that they’re not actually going to slow down,” Kokotajlo says, adding that the fear is that “they’ll just bring in some external auditors, do a bunch of safety paperwork — some of which will be genuinely good — but at the end of the day, it actually won’t slow them down very much at all.”
NYU’s Reese compared this gambit to the social media platforms’ playbook a decade ago, when companies began calling for regulatory action to get ahead of impending, less favorable laws. For the AI industry, Reese said, “the hammer may not come in this administration, but I think if there were a Democratic administration after the next election, there would be a really good possibility.”"
#AI #AISafety #OpenAI #Anthropic #Oligopolies #Antitrust #Competition #BigTech
-
"Many still question the motivations of AI’s leaders. The larger tech industry has spent years getting ahead of regulation by lobbying for its own preferred rules or promising self-regulation. Big platforms have proposed policies that could hit smaller competitors harder, using altruistic language to justify self-serving goals. They’ve been accused of safety-washing, or making meaningless changes that give the false impression of actual safeguards. It’s no surprise people are concerned this will happen in the AI industry as well, particularly since AI labs’ voluntary safety frameworks have been criticized for years.
Several sources believe that concerns of safety-washing are valid. “There’s a serious concern that they’re not actually going to slow down,” Kokotajlo says, adding that the fear is that “they’ll just bring in some external auditors, do a bunch of safety paperwork — some of which will be genuinely good — but at the end of the day, it actually won’t slow them down very much at all.”
NYU’s Reese compared this gambit to the social media platforms’ playbook a decade ago, when companies began calling for regulatory action to get ahead of impending, less favorable laws. For the AI industry, Reese said, “the hammer may not come in this administration, but I think if there were a Democratic administration after the next election, there would be a really good possibility.”"
#AI #AISafety #OpenAI #Anthropic #Oligopolies #Antitrust #Competition #BigTech
-
"Many still question the motivations of AI’s leaders. The larger tech industry has spent years getting ahead of regulation by lobbying for its own preferred rules or promising self-regulation. Big platforms have proposed policies that could hit smaller competitors harder, using altruistic language to justify self-serving goals. They’ve been accused of safety-washing, or making meaningless changes that give the false impression of actual safeguards. It’s no surprise people are concerned this will happen in the AI industry as well, particularly since AI labs’ voluntary safety frameworks have been criticized for years.
Several sources believe that concerns of safety-washing are valid. “There’s a serious concern that they’re not actually going to slow down,” Kokotajlo says, adding that the fear is that “they’ll just bring in some external auditors, do a bunch of safety paperwork — some of which will be genuinely good — but at the end of the day, it actually won’t slow them down very much at all.”
NYU’s Reese compared this gambit to the social media platforms’ playbook a decade ago, when companies began calling for regulatory action to get ahead of impending, less favorable laws. For the AI industry, Reese said, “the hammer may not come in this administration, but I think if there were a Democratic administration after the next election, there would be a really good possibility.”"
#AI #AISafety #OpenAI #Anthropic #Oligopolies #Antitrust #Competition #BigTech
-
"Many still question the motivations of AI’s leaders. The larger tech industry has spent years getting ahead of regulation by lobbying for its own preferred rules or promising self-regulation. Big platforms have proposed policies that could hit smaller competitors harder, using altruistic language to justify self-serving goals. They’ve been accused of safety-washing, or making meaningless changes that give the false impression of actual safeguards. It’s no surprise people are concerned this will happen in the AI industry as well, particularly since AI labs’ voluntary safety frameworks have been criticized for years.
Several sources believe that concerns of safety-washing are valid. “There’s a serious concern that they’re not actually going to slow down,” Kokotajlo says, adding that the fear is that “they’ll just bring in some external auditors, do a bunch of safety paperwork — some of which will be genuinely good — but at the end of the day, it actually won’t slow them down very much at all.”
NYU’s Reese compared this gambit to the social media platforms’ playbook a decade ago, when companies began calling for regulatory action to get ahead of impending, less favorable laws. For the AI industry, Reese said, “the hammer may not come in this administration, but I think if there were a Democratic administration after the next election, there would be a really good possibility.”"
#AI #AISafety #OpenAI #Anthropic #Oligopolies #Antitrust #Competition #BigTech
-
The prevailing skepticism from many anti-AI people (who I would count myself among) about AI safety discourse, and especially the distant and extreme AI risks, is so strange to me.
“AI is dangerous, but unfortunately the AI companies agree that AI is dangerous, and we can’t trust them, so it must be perfectly fine actually”—that seems to be the underlying logic.
-
https://www.europesays.com/people/228361/ Satya Nadella warns AI must stay under human control in internal memo | World News #AICodeOfConduct #AIRisks #AISafety #AISafetyRules #ArtificialIntelligence #FrontierAI #HumanControl #MicrosoftAI #MICROSOFTAICODEOFCONDUCT #SatyaNadella
-
https://www.europesays.com/people/228229/ AI CEOs Predictably Suck Up to Trump Amid ‘AI Pause’ Debate #AI #AISafety #DonaldTrump #JensenHuang #Politics #SamAltman
-
Jensen Huang, Donald Trump Denounce AI Doomerism at All-in Summit
Nvidia CEO Jensen Huang got a call from President Donald Trump in the middle of his interview to…
#NewsBeep #News #Headlines #aiadvancement #AIsafety #all-insummit #businessinsider #CA #Canada #comment #jacobcoxon #JensenHuang #lab #monday #nvidiaceo #people #PresidentDonaldTrump #Science #talk #Trump
https://www.newsbeep.com/732149/ -
Minha análise para a @TeletimeNews sobre o novo capítulo da construção do fosso de IA que está sendo escavado pelas três líderes da tecnologia nos Estados Unidos e que pode separá-las do resto do mundo se nada for feito.
https://teletime.com.br/14/09/2026/ia-o-clube-fechado-que-trabalha-em-silencio/
-
Minha análise para a @TeletimeNews sobre o novo capítulo da construção do fosso de IA que está sendo escavado pelas três líderes da tecnologia nos Estados Unidos e que pode separá-las do resto do mundo se nada for feito.
https://teletime.com.br/14/09/2026/ia-o-clube-fechado-que-trabalha-em-silencio/
-
Minha análise para a @TeletimeNews sobre o novo capítulo da construção do fosso de IA que está sendo escavado pelas três líderes da tecnologia nos Estados Unidos e que pode separá-las do resto do mundo se nada for feito.
https://teletime.com.br/14/09/2026/ia-o-clube-fechado-que-trabalha-em-silencio/
-
Minha análise para a @TeletimeNews sobre o novo capítulo da construção do fosso de IA que está sendo escavado pelas três líderes da tecnologia nos Estados Unidos e que pode separá-las do resto do mundo se nada for feito.
https://teletime.com.br/14/09/2026/ia-o-clube-fechado-que-trabalha-em-silencio/
-
Minha análise para a @TeletimeNews sobre o novo capítulo da construção do fosso de IA que está sendo escavado pelas três líderes da tecnologia nos Estados Unidos e que pode separá-las do resto do mundo se nada for feito.
https://teletime.com.br/14/09/2026/ia-o-clube-fechado-que-trabalha-em-silencio/
-
LLMs don't need to self-replicate and take over all hyperscalers/neoclouds to be a genuine danger to society, hacking just 1% of the industrial control hardware that's connected to the internet is way more than enough to cause serious casualties. That 1% capability is demonstrably here, today, even with open weight models.
-
LLMs don't need to self-replicate and take over all hyperscalers/neoclouds to be a genuine danger to society, hacking just 1% of the industrial control hardware that's connected to the internet is way more than enough to cause serious casualties. That 1% capability is demonstrably here, today, even with open weight models.
-
LLMs don't need to self-replicate and take over all hyperscalers/neoclouds to be a genuine danger to society, hacking just 1% of the industrial control hardware that's connected to the internet is way more than enough to cause serious casualties. That 1% capability is demonstrably here, today, even with open weight models.
-
AI slowdown trade hits Nvidia, SoftBank, SK Hynix as global tech stocks fall up to 10%
AI-related stocks fell across global markets on Monday after Anthropic CEO Dario Amodei called for slowing down the…
#EuropeSays #Korea #KR #SKHynix #AIdevelopment #AISafety #AIstocks #Anthropic #cloudplayers. #datacentrecompanies #globaltechstocks #NVIDIA #semiconductorstocks #SK #SKhynix #Softbank
https://www.europesays.com/korea/154269/ -
DATE: September 14, 2026 at 03:46AM
SOURCE: SOCIALPSYCHOLOGY.ORGTITLE: What to Know About Recent Dire AI Predictions and Calls for Safeguards
Source: PBS News Hour
New warnings from within the artificial intelligence industry have revived a long-running debate over whether advanced AI could threaten humanity's survival. Dario Amodei, CEO of AI company Anthropic, outlined a plan for companies and governments to ensure that increasingly capable AI models have fortified guardrails after two former Anthropic safety researchers publicly aired concerns about the existential threats AI might pose to humanity.
-------------------------------------------------
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 #AIGovernance #AISafety #AIWarnings #Anthropic #DirePredictions #Safeguards #AIethics #ExistentialRisk #TechPolicy #FutureOfAI
-
DATE: September 14, 2026 at 03:46AM
SOURCE: SOCIALPSYCHOLOGY.ORGTITLE: What to Know About Recent Dire AI Predictions and Calls for Safeguards
Source: PBS News Hour
New warnings from within the artificial intelligence industry have revived a long-running debate over whether advanced AI could threaten humanity's survival. Dario Amodei, CEO of AI company Anthropic, outlined a plan for companies and governments to ensure that increasingly capable AI models have fortified guardrails after two former Anthropic safety researchers publicly aired concerns about the existential threats AI might pose to humanity.
-------------------------------------------------
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 #AIGovernance #AISafety #AIWarnings #Anthropic #DirePredictions #Safeguards #AIethics #ExistentialRisk #TechPolicy #FutureOfAI
-
DATE: September 14, 2026 at 03:46AM
SOURCE: SOCIALPSYCHOLOGY.ORGTITLE: What to Know About Recent Dire AI Predictions and Calls for Safeguards
Source: PBS News Hour
New warnings from within the artificial intelligence industry have revived a long-running debate over whether advanced AI could threaten humanity's survival. Dario Amodei, CEO of AI company Anthropic, outlined a plan for companies and governments to ensure that increasingly capable AI models have fortified guardrails after two former Anthropic safety researchers publicly aired concerns about the existential threats AI might pose to humanity.
-------------------------------------------------
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 #AIGovernance #AISafety #AIWarnings #Anthropic #DirePredictions #Safeguards #AIethics #ExistentialRisk #TechPolicy #FutureOfAI
-
DATE: September 14, 2026 at 03:46AM
SOURCE: SOCIALPSYCHOLOGY.ORGTITLE: What to Know About Recent Dire AI Predictions and Calls for Safeguards
Source: PBS News Hour
New warnings from within the artificial intelligence industry have revived a long-running debate over whether advanced AI could threaten humanity's survival. Dario Amodei, CEO of AI company Anthropic, outlined a plan for companies and governments to ensure that increasingly capable AI models have fortified guardrails after two former Anthropic safety researchers publicly aired concerns about the existential threats AI might pose to humanity.
-------------------------------------------------
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 #AIGovernance #AISafety #AIWarnings #Anthropic #DirePredictions #Safeguards #AIethics #ExistentialRisk #TechPolicy #FutureOfAI
-
Representatives of Google, OpenAI and Anthropic are having on-going conversations about forming an AI industry standards body to establish concrete standards ... and they will have more to share “in the next months.” https://www.cnn.com/2026/09/14/tech/ai-standards-body #AI #AIStandards #Regulation #Google #OpenAI #Anthropic #IndustryStandards #AISafety #AIPolicy #AILeaders
-
Representatives of Google, OpenAI and Anthropic are having on-going conversations about forming an AI industry standards body to establish concrete standards ... and they will have more to share “in the next months.” https://www.cnn.com/2026/09/14/tech/ai-standards-body #AI #AIStandards #Regulation #Google #OpenAI #Anthropic #IndustryStandards #AISafety #AIPolicy #AILeaders
-
Representatives of Google, OpenAI and Anthropic are having on-going conversations about forming an AI industry standards body to establish concrete standards ... and they will have more to share “in the next months.” https://www.cnn.com/2026/09/14/tech/ai-standards-body #AI #AIStandards #Regulation #Google #OpenAI #Anthropic #IndustryStandards #AISafety #AIPolicy #AILeaders