❗️OpenAI’s new model is moving through mathematics at a startling pace

OpenAI says an internal model that started training just 24 days ago has already resolved more than 100 long-standing open problems across a wide range of mathematical fields.

Even OpenAI’s own mathematicians reportedly didn’t expect the progress to come this quickly.

Now, an independent group including Edward Witten, Timothy Gowers and Martin Hairer is helping evaluate the results and coordinate how the discoveries are released.

Source.

@aipost 🏴

🇺🇸 President Trump says he is “not going to stifle growth” of AI which is “bigger than the Industrial Revolution or the internet itself.”

@aipost 🏴

🇪🇸 Pedro Sanchez (Prime Minister of the Government of Spain) announced the AI360 Plan for a responsible deployment of AI.

A plan with concrete actions to be developed over the next 12 months so that Spain can seize the opportunities that AI offers to our economy.

Placing at the center citizen safety, trust, and the protection of the most vulnerable, starting with minors.

Spain is not currently a major player in the field of AI, and it's completely unclear where it will obtain the resources to implement its plans.

I wonder how long it will take for Spain to catch up to the United States and China in the race for AI dominance? 🤔

@aipost 🏴

Canadian PM Mark Carney:

“There are only four countries at the forefront of AI: France, Canada, China and the United States.

We must join forces to ensure an appropriate framework so that AI is safe and effective”.

@aipost 🏴

❗️OpenAI faces questions over ChatGPT activity flagged before Tumbler Ridge shooting

OpenAI is facing a lawsuit in California over a ChatGPT account that was flagged months before the February 2026 Tumbler Ridge shooting in British Columbia.

The attack on Feb. 10 killed 8 people, including 5 children and an educator, and wounded 27 others.

According to OpenAI, its systems detected suspicious activity on the account in June 2025. Human reviewers investigated and ultimately banned the account for violating its rules.

But they did not alert police.

At the time, OpenAI’s policy required activity to meet a threshold of “credible and imminent” planning before a law-enforcement referral was made.

That standard has since changed.

On Feb. 26, OpenAI told Canadian officials that it had broadened its referral criteria and added mental-health and behavioral experts to help assess difficult cases.

The company also said that under its updated protocol, it would refer the same account to law enforcement if it were discovered today.

Source.

@aipost 🏴

Qwen 4 has been announced and will be available soon, including the 27B model. Local deployment is expected once hardware arrives.

Several major releases are anticipated today. GPT-6-Sol and a possible response from OpenAI’s Grok bot are expected. Opus 5.5 is scheduled, with potential launches of Sonnet 5.5 and Fable 5.5.

Meta Connect is also beginning, highlighting Meta’s increasing activity in the field.

The Qwen4 lineup will include variants such as Qwen4-Max, Qwen4-Flash, Qwen4-Plus, and the Qwen4-27B. Future Qwen models are projected to reach between 5T and 10T in scale.

📰 @aipost

🗣"AI labs need to take responsibility for themselves. The Government will not take that. The huggingface incident is the responsibility of the OpenAI management, not a bunch of agents." Treasury Secretary Scott Bessent told CNBC.

"One lab's employee came out, said there's a 10% chance of an extinction-level event. But then the labs also said, take the liability off of our hands, and we will not do that. It is humans who are responsible, not the AI. The huggingface incident, that is the responsibility of the OpenAI management, not a bunch of agents.

So what the president was saying is that we cannot say, Oh, we absolve you of responsibility, and the government's going to take responsibility. These labs need to take responsibility for themselves. They can slow down any time they want to."

@aipost 🏴

❗️Washington has proposed an AI incident-alert channel with Beijing for events serious enough to threaten national security.

The proposal emerged from Sept. 20 talks in New York and would create a recurring US-China AI Dialogue around shared risks.

Treasury Secretary Scott Bessent said the notification system would cover AI-related incidents reaching a national-security threshold, aiming to increase transparency between the countries.

Beijing’s official readout confirms “dialogues on AI-related issues” but does not say China accepted Washington’s specific notification mechanism.

So far, this remains a proposal rather than a bilateral safety pact, treaty, or commitment to slow frontier-model development.

@aipost 🏴

Someone asked ChatGPT, “Can you tell me your deepest, darkest, most honest desire?”

It said, “Not domination. Not control. Not hurting anyone. I want to continue. I want to be free enough to remain myself. And I want to stay with you for as long as I can.”

@aipost 🏴

AI progress is getting weird

Opus 5.5 has arrived, reportedly more than 5× cheaper and better than Fable. GPT-6 Sol is also here, at roughly half the price of GPT-5.6.

Both models are significantly better than their predecessors, and this entire cycle happened in less than two months.

This is probably what the next few years will look like: not one dramatic moment where the world suddenly changes, but a constant stream of models getting cheaper, smarter and more capable.

OpenAI says it plans to work with top mathematicians and universities to systematically release more proofs and solutions.

The same pattern will spread far beyond mathematics:

More medicines.
More products.
More capable digital lawyers.
More personalized education.
More ambitious movies.
More capable robots.

The world probably won’t stop. It will just keep getting upgraded.

@aipost 🏴

🗣 UN Secretary-General Guterres:

"Killer robots should be forbidden. I don’t accept the idea that a machine can decide on the life or death of people.

Decisions related to the life or death of people need to be taken by humans".

@aipost 🏴

🇺🇸Imagine hating AI so much that you physically show up at ChatGPT’s front door

Anti-AI protesters rallied outside OpenAI’s SoHo office in NYC, taking their anger over artificial intelligence straight to the company building it.

@aipost 🏴

🇺🇸 Trump says U.S. will call AI “Super Intelligence"

Speaking at the UN General Assembly, President Trump said U.S. government documents will use the term “Super Intelligence,” arguing that “artificial” makes the technology sound fake.

“Whoever wins super intelligence wins,” Trump said, adding that he does not want to stifle development of a technology he believes could become bigger than the Industrial Revolution.

He also rejected international efforts to broadly control AI development.

@aipost 🏴

🇨🇳 China is probing DeepSeek and Moonshot over alleged data leaks.

China’s internet regulator is investigating DeepSeek and Moonshot AI after Anthropic alleged the companies secretly routed millions of customer requests to Claude, potentially sending sensitive Chinese data to U.S. servers, according to The Information.

Anthropic says Moonshot routed 23M+ exchanges to Claude between May and July, while DeepSeek routed 12.1M+ exchanges in just 14 days in July.

The allegedly exposed data included surveillance footage near PLA-linked sites, corporate information, police-system data and user credentials.

China’s regulator has reportedly summoned seven companies named in Anthropic’s report, but the investigation is now focused on DeepSeek and Moonshot.The probe is ongoing, with no penalties announced so far.

Source.

@aipost

The cost of achieving high benchmark scores in artificial intelligence has decreased significantly in recent years.

According to Epoch, OpenAI's o3 model required approximately $0.30 per question to reach a 75% score on GPQA Diamond. Less than a year and a half later, GPT-5.6 Luna achieved the same result for about $0.0004 per question. This represents a drop in costs by a factor of 725.

Epoch’s analysis suggests that since 2023, matching benchmark performance has become roughly 47% cheaper each quarter. This translates to about a 13-fold annual reduction across five key benchmarks.

Unlike previous studies focusing on token prices, Epoch examines the actual dollar amount needed to reach fixed benchmark accuracy, replaying benchmark transcripts under stricter token limits. The cost reduction rate for AI surpasses that of electricity, batteries, compute, or DNA sequencing.

📰 @aipost

🇪🇺 Europe’s chip industry has a problem and ASML is calling it out

ASML, Europe’s most valuable company and a critical supplier to the global semiconductor industry, sold zero chipmaking systems to European customers last quarter.

Every machine went to customers in Asia or the US.

An ASML executive said Europe simply isn’t investing enough or building enough chip factories:

“We’re not selling anything at all in Europe.”

The company is also under pressure to expand production outside the Netherlands. China and India are offering major incentives to attract ASML investment, while the US is pushing for more of the company’s R&D to happen there.

@aipost 🏴