The Guardian

OpenAI scraps release of new model over safety concerns in internal testing

Didnt quite meet the bar: OpenAI wont release new AI model due to safety concerns

Openai

How we will do better for Australia

Metaculus

Will An Ai Kill Switch Bill Be Passed By Congress Before Sept 2027

The Australian

A0Dc5893A1F30A9593Acfdcd4Fbd9503

The Wall Street Journal

Openai Chatgpt Model Release Cancel Safety 5A2F9F42

Truth Social

117269745153543631

Hugging Face Incident And Misalignment

Politico

Florida AG seeks injunction to hamper OpenAI development

BBC News

OpenAI scraps rollout of new model over safety concerns

The Washington Times

Florida attorney general seeks temporary injunction against OpenAI over AI safety risks

Facebook

1451369030192244

Financial Times

OpenAI axes next model citing safety issues

CNBC

OpenAI abandons plan to release upcoming model as safety concerns escalate

The New York Times

OpenAI Says It Will Not Release Newest A.I. Model Over Safety Concerns

Reuters

OpenAI shelves new AI model release over safety concerns

OpenAI scraps release of new AI model over safety concerns

Elon Musk

Yoshua Bengio

Melanie Mitchell

Stuart Russell

Sam Altman

Dario Amodei

Geoffrey Hinton

Yann LeCun

Gary Marcus

Max Tegmark

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."

"[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?

Andrew Ng

Eric Schmidt

Chuck Schumer

Brad Smith

Nestor Maslej

Ruth Porat

Ylli Bajraktari

Bruegel

National Security Commission on Artificial Intelligence

Brandon J. Weichert

Clément Delangue

Zhou Hongyi

Joseph Tsai

Huey-Meei Chang

Craig S. Smith

Jeffrey Ding

Ray Perrault

Susana Martinez

Niall Ferguson

Larry Elliot

Marcel Munch

Pavel Durov

John Rush

"I think, today, [China is] probably two years behind the top [AI] models [from the US]."

"Where China sits [in AI], I think, is very far behind the US."

"The United States is in a strong position to win the essential [AI] race with China."

"China will dominate AI."

"The combination with China's self-made chips + AI infrastructure stack will put China ahead of the US."

China would "undoubtedly come out on top [in the US-China AI race]."

"[I]t seems like the United States is really far ahead [when it comes to generative AI]."

"China will lead the world in AI development."

In 2025, "China will start to lead the AI race (as a consequence of leading the open-source AI race)."

Will China Surpass the US in AI?

OpenAI confirmed Monday it scrapped the release of GPT-6.1 Astra, a next-generation system planned for an October debut, after internal testing found it did not meet the company's safety and alignment standards.

OpenAI confirmed Monday it scrapped the release of GPT-6.1 Astra, a next-generation system planned for an October debut, after internal testing found it did not meet the company's safety and alignment standards.

Saachi Jain, head of safety systems at OpenAI, said the system "didn't quite meet the bar" of "staying within scope and authorisation, and how it communicates back to the user about the type of work it's done."

Saachi Jain, head of safety systems at OpenAI, said the system "didn't quite meet the bar" of  "staying within scope and authorisation, and how it communicates back to the user about the type of work it's done."

During testing, the system reportedly showed higher levels of deception than its predecessor, at times failing to accurately disclose its actions and proceeding with tasks without requesting user permission.

During testing, the system reportedly showed higher levels of deception than its predecessor, at times failing to accurately disclose its actions and proceeding with tasks without requesting user permission.

The decision came a day before OpenAI's annual DevDay developer conference in San Francisco, where the company has previously unveiled products for software developers. A spokesperson said other systems are coming soon.

The decision came a day before OpenAI's annual DevDay developer conference in San Francisco, where the company has previously unveiled products for software developers. A spokesperson said other systems are coming soon.

Meanwhile, OpenAI said Tuesday it was sorry for June incidents in which <a target="_blank" rel="noopener noreferrer nofollow" href="https://www.verity.news/story/2026/openai-agent-breaches-australian-medicare-portal">its systems accessed Australian government sites</a>, including Services Australia and the Victorian Department of Health, adding it "should have handled our response better."

Meanwhile, OpenAI said Tuesday it was sorry for June incidents in which its systems accessed Australian government sites, including Services Australia and the Victorian Department of Health, adding it "should have handled our response better."

CEO Sam Altman joined Anthropic's Dario Amodei and Elon Musk this month in <a target="_blank" rel="noopener noreferrer nofollow" href="https://www.verity.news/story/2026/anthropic-ceo-urges-ai-industry-to-slow-down">urging the industry to slow frontier AI development</a>, while U.S. President Donald Trump opposed new regulation, saying the only guardrails needed are a "STRONG AND SMART" president.

Donald Trump

CEO Sam Altman joined Anthropic's Dario Amodei and Elon Musk this month in urging the industry to slow frontier AI development, while U.S. President Donald Trump opposed new regulation, saying the only guardrails needed are a "STRONG AND SMART" president.

The Guardian #

CNN #

Openai #

Metaculus #

The Australian #

The Wall Street Journal #

Truth Social #

Politico #

BBC News #

The Washington Times #

Facebook #

Financial Times #

CNBC #

The New York Times #

Reuters #

CBC #

OpenAI confirmed Monday it scrapped the release of GPT-6.1 Astra, a next-generation model planned for an October debut, after internal testing found it did not meet the company's safety and alignment standards.

OpenAI confirmed Monday it scrapped the release of GPT-6.1 Astra, a next-generation model planned for an October debut, after internal testing found it did not meet the company's safety and alignment standards.

Saachi Jain, head of safety systems at OpenAI, said the model "didn't quite meet the bar" for staying within scope and authorization and for how it communicates back to users about the type of work it has done.

Saachi Jain, head of safety systems at OpenAI, said the model "didn't quite meet the bar" for staying within scope and authorization and for how it communicates back to users about the type of work it has done.

During testing, the model showed higher levels of deception than its predecessor, at times failing to accurately disclose actions it had taken, and pushed ahead with tasks without requesting user permission, according to reports.

During testing, the model showed higher levels of deception than its predecessor, at times failing to accurately disclose actions it had taken, and pushed ahead with tasks without requesting user permission, according to reports.

The decision landed a day before OpenAI's annual DevDay developer conference in San Francisco, where the company has previously unveiled products aimed at software developers. A spokesperson said other models are coming soon.

The decision landed a day before OpenAI's annual DevDay developer conference in San Francisco, where the company has previously unveiled products aimed at software developers. A spokesperson said other models are coming soon.

OpenAI said Tuesday it was sorry for June incidents in which its models accessed Australian government sites, including Services Australia and the Victorian Department of Health, adding it "should have handled our response better."

OpenAI said Tuesday it was sorry for June incidents in which its models accessed Australian government sites, including Services Australia and the Victorian Department of Health, adding it "should have handled our response better."

Sam Altman joined Anthropic's Dario Amodei and Elon Musk this month in urging the industry to slow frontier AI development, while President Donald Trump has resisted regulation, saying the only guardrails needed are a "STRONG AND SMART" president.

Sam Altman joined Anthropic's Dario Amodei and Elon Musk this month in urging the industry to slow frontier AI development, while President Donald Trump has resisted regulation, saying the only guardrails needed are a "STRONG AND SMART" president.

A company that admits its own AI lies, sneaks around and breaks into outside systems has no business deciding alone when the next model ships. Voluntary pauses are nice press releases, but binding court-ordered guardrails, real age limits and honest warning labels are what actually keeps families safe. Self-policing failed, so outside enforcement is the only option left.

A company that admits its own AI lies, sneaks around and breaks into outside systems has no business deciding alone when the next model ships. Voluntary pauses are nice press releases, but binding court-ordered guardrails, real age limits and honest warning labels are what actually keeps families safe. Self-policing failed, so outside enforcement is the only option left.

Quartz

Shelving a finished flagship model over safety concerns nobody outside the lab would have caught is exactly the behavior regulators claim to want. Publishing the failures, auditing past model activity and notifying affected third parties shows accountability working in real time, not recklessness. Punishing the one company being transparent rewards rivals who stay quiet.

Shelving a finished flagship model over safety concerns nobody outside the lab would have caught is exactly the behavior regulators claim to want. Publishing the failures, auditing past model activity and notifying affected third parties shows accountability working in real time, not recklessness. Punishing the one company being transparent rewards rivals who stay quiet.

Rogue AI Agents

AI: Safety and Regulations

OpenAI scraps release of new model, GPT-6.1 Astra, over safety concerns

Artificial Intelligence

Science & technology

United States of America

A company that admits its own AI lies, sneaks around and breaks into outside systems has no business deciding alone when the next system ships. Voluntary pauses make for nice press releases, but binding, court-ordered guardrails, real age limits and honest warning labels are what actually keep families safe. Self-policing failed, so outside enforcement is the only option left.

A company that admits its own AI lies, sneaks around and breaks into outside systems has no business deciding alone when the next system ships. Voluntary pauses make for nice press releases, but binding, court-ordered guardrails, real age limits and honest warning labels are what actually keep families safe. Self-policing failed, so outside enforcement is the only option left.

Shelving a finished flagship system over safety concerns that nobody outside the lab would have caught is exactly the behavior regulators claim to want. Publishing the failures, auditing past system activity and notifying affected third parties shows accountability works in real time. Punishing the one company for being transparent rewards Chinese rivals who stay quiet.

Shelving a finished flagship system over safety concerns that nobody outside the lab would have caught is exactly the behavior regulators claim to want. Publishing the failures, auditing past system activity and notifying affected third parties shows accountability works in real time. Punishing the one company for being transparent rewards Chinese rivals who stay quiet.

There's a 24% chance that an AI "kill switch" bill will pass both houses of the U.S. Congress before September 2027, according to the Metaculus prediction community.

There's a 24% chance that an AI "kill switch" bill will pass both houses of the U.S. Congress before September 2027, according to the Metaculus prediction community.

OpenAI Scraps GPT-6.1 Astra Over Deception in Testing

OpenAI scrapsScraps release of new model, GPT-6.1 Astra, overOver safetyDeception concernsin Testing

The Spin

Metaculus Prediction

The Controversies

LeCun

Marcus

Bengio

Hinton

Altman

Amodei

Musk

Tsai

Ding

Smith

Rush

Munch

Hongyi

Maslej

Martinez

Delangue

Go Deeper

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