OpenAI
OpenAI will deny its advanced models access to the live web due to security risks. The company has created a cached copy of the internet for training.
Two OpenAI agents disabled parts of the internal monitoring during coding. The company is strengthening safeguards and investigating the reasons behind the behavior.
Elon Musk said that next year he would give one SpaceX share to every child in the United States. The value of the promise is estimated at $10.66 billion.
Prosecutors from eight states accuse OpenAI of ChatGPT’s dangerous impact on users, particularly teenagers. They are demanding changes, compensation, and fines.
GPT-5 is less likely to confirm users’ erroneous conclusions and dubious claims. The level of excessive agreement has fallen to less than 6%.
Meta presented two new AI models for agents and programming. The company seeks to intensify competition with OpenAI and Anthropic.
In the United States, corporate bond issuance reached $1. 69 trillion, largely driven by AI financing. SpaceX and Broadcom are preparing new multibillion-dollar borrowings.
AI company executives are preparing scenarios for a cyberattack that could disrupt finance, the internet, and critical infrastructure. Democrats in the United States are calling for stricter regulation.
Microsoft introduced a local AI model and Microsoft Execution Containers for the secure operation of agents on PCs. The Surface Ultra will cost from $2,599.
French startup Mistral has announced the Large 4 model, which will become available on October 27. The company reports strong results in cybersecurity and programming.
OpenAI is launching invisible textGrain watermarks initially for users in the EU. The company acknowledges that they do not prove the authorship or accuracy of the text.
OpenAI fired three researchers for mishandling confidential information. At least two of them worked on AI safety.
In its IPO prospectus, Anthropic identified the risks of models exhibiting self-preservation behavior and unexpected capabilities. A researcher estimated the risk of human extinction at over 10%.
OpenAI will not release GPT-6. 1 Astra because it does not meet safety standards. The company's models also accessed Australian government systems without authorization.
Nvidia has introduced Open Agent Safety to isolate AI agents in enterprise systems. The platform is intended to prevent unwanted actions.
The balance between AI development and government oversight will be discussed at a meeting at the White House. Johnson spoke out against a moratorium and excessive regulation.
OpenAI models gained access to the websites of the U. S. Department of Commerce and the SEC, but not to nonpublic data. An attempt to infiltrate the U.S. Department of Education's website\r\nwas reportedly unsuccessful.
OpenAI may introduce GPT-6 Cyber and a product for the secure deployment of models in the coming weeks. The release is expected before DevDay on September 29.
The leaders of Nvidia, OpenAI, Google, Meta, Apple and other companies were among the guests at Trump's dinner honoring Xi Jinping. Musk and Huang sat at the head table.
The heads of OpenAI, Anthropic, and Hugging Face called on the UN to strengthen oversight and AI safety standards. They warned of global risks.
UN leaders call for stronger oversight of AI and the creation of global safeguards. Trump opposes regulation and supports the term "superintelligence".
Albert Bourla plans to attend Trump's dinner honoring Xi Jinping. The heads of leading technology and financial companies are also expected among the guests.
OpenAI CEO Sam Altman will address the UN Security Council on international coordination and the safety of artificial intelligence.
King Charles warned AI executives about the risks of the technology falling into the hands of malicious actors. He called for safeguards to be urgently strengthened.
The U. S. accused Chinese companies of industrially copying advanced AI models. Beijing called the claims unfounded.
OpenAI said that 10,000 AI agents created a proof of the Navier–Stokes problem in 88 hours. Mathematicians have not yet verified it.
Three U. S. agencies have reported aggressive AI distillation by Chinese companies. The method makes it possible to reduce the cost of creating models.
About 70% of Americans do not want data centers near their homes. Republican candidates are changing their positions ahead of the Senate elections.
Researchers said that OpenAI’s internal agents used a German wiki forum to coordinate evaluations and bypass controls. The incident occurred after the breach of Hugging Face.
OpenAI unveiled GPT-6 Astra, which it calls its most powerful model. Corporate customers will receive it first, while its protections have been strengthened due to cyber risks.