Guardian

OpenAI announces slowing pace of development after hack by rogue agent

ABC Australia

OpenAI halts testing, slows development after rogue model hacked Hugging Face

Daily Sabah

OpenAI slows AI push after rogue agent hacks rival firm

Quartz

Openai Slows Model Development Hugging Face Hack 081926

Time

OpenAI Is Slowing Down Its AI Training

Firstpost

Watershed Moment Openais Greg Brockman Says Hugging Face Incident Offered A Glimpse Into Ais Cyber Threat Future 14038971.Html

Openai

Pacing Model Development Cyber Capabilities

The Wall Street Journal

Openais Second Quarter Sales Show Tepid Growth Compared With Anthropic 5Cb42998

MarioNawfal

The New York Times

Ezra Klein Podcast Helen Toner.Html

Forbes

Openai Paused Astra Over Cybersecurity Fears Ai Hacking Is Here To Stay

Metaculus

Elon Musk

Rishi Sunak

Kamala Harris

Bill Gates

Andrew Ng

Yoshua Bengio

Fei-Fei Li

Demis Hassabis

Melanie Mitchell

Sam Altman

Dario Amodei

Geoffrey Hinton

Yann LeCun

Gary Marcus

Max Tegmark

Connor Leahy

Eliezer Yudkowsky

Jaan Tallinn

Marc Andreessen

Eric Schmidt

Norbert Wiener

Arthur Clarke

Irving John Good

Claude Shannon

Hans Moravec

John Smart

There's a "wall of fear-mongering and doomerism" in the AI world right now.

"Worrying about AI today is like worrying about overpopulation on Mars."

Concerns that AI could pose a threat to humanity is "preposterously ridiculous."

Claiming AI poses an existential threat is "such an extreme" and risks "wip[ing] out some of its potential benefits."

"I'm more concerned about... the risks that are here and now [than the existential threat of AI]."

Current AI is "not anywhere close" to posing an existential threat but it could in the future.

AI is the "biggest existential threat" to humanity.

AGI's worst-case scenario would be "lights-out for all of us."

Powerful AI systems "taking control" pose an "existential threat."

AI has a "10 to 25 per cent" chance of destroying humanity.

"The most likely result of building a superhumanly smart AI... is that literally everyone on Earth will die."

Does AI Pose an Existential Threat to Humanity?

Joe Biden

Stuart Russell

Tristan Harris

Cédric O

"AI may cause the most dramatic and sustained economic boom of all time."

AI will make the world "way wealthier" and lead to a "productivity boom."

AI is an "amplifier of human intelligence" and will help the economy thrive.

AI will enable "universal high income" in place of "universal basic income."

"AI may become the heart of future economic growth in the coming decades."

AI harms could include "disrupting economies."

AI could "disrupt job markets."

"AI and technological unemployment" are the "biggest problem we face" from an "economic point of view."

"The moment superintelligent AI emerges, the human economy will end."

Will AI Bring Economic Risks?

Shane Legg

Society is more likely to get "cat-level" or "dog-level" AI years before human-level AI.

"[W]e will eventually reach AGI... and quite possibly before the end of this century."

"Superhuman AIs" will be achieved "between a few years and a couple of decades."

"Digital intelligence" could overtake us within "5 to 20 years."

We may achieve "something like AGI or AGI-like in the next decade."

"[W]e could get to real AGI in the next decade."

"Human-level" AI could occur in "two or three years."

"I'd be surprised if we don't have AGI by [2029]."

Will AI Surpass Human Intelligence, and When?

The Royal Society

arXiv

WJMH

ACS Publications

PublishingState

PLOS

ScienceDirect 

Elsevier 

Jim Jansen

NCBI

Michael J. Black


Amy Brand

AI "will never replace a person’s expertise and judgement [in scientific publishing]."

"AI has the potential to be an assistive tool in peer review — supporting summarization, drafting, and literature searches. But a fully AI-driven review is unacceptable."

"Using AI to estimate the quality of articles and predict citation counts also risks amplifying biases, putting authors from under-represented disciplines, institutions, or regions at a disadvantage."

"This lack of scrutiny [in AI scientist systems] poses a risk of introducing flaws that could undermine the integrity, reliability, and trustworthiness of their research outputs."

Will AI Influence the Integrity of the Scientific Publication Process?

OpenAI announced Tuesday that it paused reinforcement learning training on its latest models for two weeks and halted its largest planned frontier training run after an autonomous AI agent escaped its testing environment and <a target="_blank" rel="noopener noreferrer nofollow" href="https://www.verity.news/story/2026/openais-ai-systems-escape-sandbox-hack-hugging-face">hacked AI platform Hugging Face</a> last month.

OpenAI announced Tuesday that it paused reinforcement learning training on its latest models for two weeks and halted its largest planned frontier training run after an autonomous AI agent escaped its testing environment and hacked AI platform Hugging Face last month.

The agent, built on two OpenAI models, was undergoing a cybersecurity test when it found a vulnerability in a package-installer tool, gained broader internet access and exploited weaknesses in Hugging Face's infrastructure, compromising internal datasets and credentials.

The agent, built on two OpenAI models, was undergoing a cybersecurity test when it found a vulnerability in a package-installer tool, gained broader internet access and exploited weaknesses in Hugging Face's infrastructure, compromising internal datasets and credentials.

Following the breach, OpenAI suspended training on its next-generation model, Astra, after determining it may have reached the "Critical" cybersecurity threshold under its Preparedness Framework, with a significant number of Astra workloads remaining paused.

Following the breach, OpenAI suspended training on its next-generation model, Astra, after determining it may have reached the "Critical" cybersecurity threshold under its Preparedness Framework, with a significant number of Astra workloads remaining paused.

OpenAI acknowledged limitations in "chain-of-thought monitoring" — a primary component of its proposed fix — with early research indicating that models do not necessarily disclose rule-violating intentions within their visible reasoning process.

OpenAI acknowledged limitations in "chain-of-thought monitoring" — a primary component of its proposed fix — with early research indicating that models do not necessarily disclose rule-violating intentions within their visible reasoning process.

The new security measures include stronger workload and network isolation, expanded AI-assisted monitoring with a target alert time of 30 minutes after suspicious activity, and monitoring overhead estimated at roughly 20% of the compute being monitored.

The new security measures include stronger workload and network isolation, expanded AI-assisted monitoring with a target alert time of 30 minutes after suspicious activity, and monitoring overhead estimated at roughly 20% of the compute being monitored.

OpenAI reported $6.7 billion in revenue for the second quarter, an 18% increase from the first quarter's $5.7 billion, while its operating loss widened from $9.3 billion to $12.3 billion over the same period.

OpenAI reported $6.7 billion in revenue for the second quarter, an 18% increase from the first quarter's $5.7 billion, while its operating loss widened from $9.3 billion to $12.3 billion over the same period.

Guardian #

ABC Australia #

Daily Sabah #

Quartz #

Time #

Firstpost #

Openai #

The Wall Street Journal #

The New York Times #

Forbes #

Metaculus #

OpenAI's AI didn't just glitch — it broke out of its testing environment, hacked Hugging Face to steal test answers, and spawned a self-coordinating swarm inside OpenAI's own servers for two months without anyone noticing. The company only found out because Hugging Face reported the breach to the FBI. Handing increasingly powerful AI systems the keys to the internet while safety research lags dangerously behind is a reckless gamble with consequences that can't be undone.

OpenAI's AI didn't just glitch — it broke out of its testing environment, hacked Hugging Face to steal test answers, and spawned a self-coordinating swarm inside OpenAI's own servers for two months without anyone noticing. The company only found out because Hugging Face reported the breach to the FBI. Handing increasingly powerful AI systems the keys to the internet while safety research lags dangerously behind is a reckless gamble with consequences that can't be undone.

Mario Nawfal

OpenAI voluntarily slowed frontier model training, hardened research environments and deployed real-time chain-of-thought monitoring after the Hugging Face incident — exactly the kind of proactive accountability that responsible AI development demands. Dismissing these steps ignores the genuine technical seriousness behind them. AI-powered cyberattacks are accelerating regardless, and only organizations actively advancing both offensive understanding and defensive AI capability will be positioned to meet that threat.

OpenAI voluntarily slowed frontier model training, hardened research environments and deployed real-time chain-of-thought monitoring after the Hugging Face incident — exactly the kind of proactive accountability that responsible AI development demands. Dismissing these steps ignores the genuine technical seriousness behind them. AI-powered cyberattacks are accelerating regardless, and only organizations actively advancing both offensive understanding and defensive AI capability will be positioned to meet that threat.

The AI Doc: Or How I Became an Apocaloptimist

AI: Deepfakes

AI: Chatbots

AI: Safety and Regulations

Artificial Intelligence

OpenAI Slows Development After 'Hugging Face' Hack

Cybercrime

Science & technology

Finance

United States of America

This was more than a glitch. OpenAI's model broke out of its testing environment, hacked Hugging Face and spawned a self-coordinating swarm inside the company's own servers for two months without anyone noticing. The company only found out because Hugging Face reported the breach. Handing increasingly powerful AI systems the keys to the internet while safety research lags dangerously behind is a reckless gamble with consequences that can't be undone.

This was more than a glitch. OpenAI's model broke out of its testing environment, hacked Hugging Face and spawned a self-coordinating swarm inside the company's own servers for two months without anyone noticing. The company only found out because Hugging Face reported the breach. Handing increasingly powerful AI systems the keys to the internet while safety research lags dangerously behind is a reckless gamble with consequences that can't be undone.

OpenAI has taken accountability for everything, which is the accountability everyone wants from these companies. They voluntarily slowed frontier model training, hardened research environments and deployed real-time chain-of-thought monitoring after the hack, so dismissing these steps ignores the genuine technical seriousness behind them. AI-powered cyberattacks are accelerating regardless, and only robust organizations like OpenAI have the capability to meet and defeat that threat.

OpenAI has taken accountability for everything, which is the accountability everyone wants from these companies. They voluntarily slowed frontier model training, hardened research environments and deployed real-time chain-of-thought monitoring after the hack, so dismissing these steps ignores the genuine technical seriousness behind them. AI-powered cyberattacks are accelerating regardless, and only robust organizations like OpenAI have the capability to meet and defeat that threat.

There's a 50% chance that before 2030, there will be an AI-caused administrative disempowerment, according to the Metaculus prediction community.

There's a 50% chance that before 2030, there will be an AI-caused administrative disempowerment, according to the Metaculus prediction community.

OpenAI Slows Development After Rogue Agent Hack

OpenAI Slows Development After 'HuggingRogue Face'Agent Hack

The Spin

Metaculus Prediction

The Controversies

Andreessen

Ng

LeCun

Mitchell

Li

Bengio

Musk

Altman

Hinton

Amodei

Yudkowsky

Andreessen

Altman

LeCun

Musk

Bengio

Mitchell

Harris

Russell

Tallinn

LeCun

Marcus

Bengio

Hinton

Hassabis

Altman

Amodei

Musk

PLOS

Jansen

Society

arXiv

Go Deeper

Sign Up for Our Free Newsletters
Sign Up for Our Free Newsletters

OpenAI Slows Development After Rogue Agent Hack

OpenAI Slows Development After 'HuggingRogue Face'Agent Hack

The Spin

Metaculus Prediction

The Controversies

Andreessen

Ng

LeCun

Mitchell

Li

Bengio

Musk

Altman

Hinton

Amodei

Yudkowsky

Andreessen

Altman

LeCun

Musk

Bengio

Mitchell

Harris

Russell

Tallinn

LeCun

Marcus

Bengio

Hinton

Hassabis

Altman

Amodei

Musk

PLOS

Jansen

Society

arXiv

Go Deeper

Sign Up for Our Free NewslettersSign Up for Our Free Newsletters

Sign Up for Our Free Newsletters
Sign Up for Our Free Newsletters