World leader accuses AI agent of hacking into government website outside the US

Coinbase CEO Brian Armstrong sees a huge opening for the next generation of tech founders, and his company is betting on a new San Francisco academy to help find them.

“Opportunity for innovation is massive right now, especially with AI + crypto,” Armstrong wrote on X. “There’s never been a better time to be a builder.”

His post came in response to the launch of the Horowitz Andreessen Academy, which aims to recruit recent high school graduates and teach them to build companies and products through hands-on projects. Coinbase is one of its 10 founding partners, alongside Anthropic, Google, Meta, Nvidia and OpenAI, among others.

Academy CEO Gagan Biyani said students will work on their own projects, take classes and hear from tech leaders. The school has raised $42 million in funding led by Andreessen Horowitz and says its network includes more than 50 hiring partners and 200 speakers and mentors.

Applications are open for a one-year, tuition-free founding fellowship. Biyani said the academy eventually hopes to offer a two-year program that charges tuition, pending regulatory approval.

Armstrong said Coinbase joined the effort to “help find and grow the next wave of world changing founders.”

Sen. Bernie Sanders went after President Donald Trump’s star-studded dinner with Chinese President Xi Jinping, accusing the tech titans on the guest list of getting richer while Americans face an uncertain future with artificial intelligence.

“Trump invited 10 Big Tech Oligarchs to his private dinner with President Xi of China,” the Vermont independent wrote on X.

Sanders claimed the 10 guests have nearly doubled their combined wealth since Trump’s election and are now worth almost $2 trillion. He singled out Elon Musk, claiming the Tesla CEO’s fortune grew by $658 billion, followed by Michael Dell at $141 billion and Nvidia CEO Jensen Huang at $70 billion. Sanders did not provide a source for his figures in the post.

Then came his warning about who could win big as AI reshapes everyday life.

“The function of AI must be to improve life for all humanity, not just make a handful of billionaires even richer,” Sanders wrote.

Musk, Dell and Huang are among the tech leaders U.S. officials said would attend the White House state dinner. AI is expected to be one of the issues discussed during Xi’s visit, along with trade and the flow of rare earth minerals.

Sen. Mark Warner, D-Va., announced plans to introduce a Senate bill requiring mandatory pre-release safety testing for artificial intelligence models.

Basic cyber standards and safety protocols for AI test beds are an urgent necessity supported by major U.S. corporations, not just tech companies, Warner told reporters Thursday.

"We can debate about data centers, kids safety, AI liability, but let's at least do a pre-model release safety testing," Warner said. "Let's at least make sure we put in cyber standards around the test beds, which have been kind of jumped over. Let's make sure there was at least a safety protocol. It would be crazy not to do that minimum."

San Francisco-based startup Trebellar has raised $18 million in a Series A funding round led by Blossom Capital, aiming to bring a stock to Wall Street to help modernize corporate real estate management with artificial intelligence.

Enterprise real estate teams manage massive amounts of operational data, but it is often fragmented across disconnected spreadsheets, static tools, and expensive consultancies.

Trebellar provides an AI-native platform that unifies lease, occupancy, and market data into a single system of record. Its AI agents can run predictive analytics, generate instant reports, and recommend agile cost-saving measures tailored to a company's portfolio.

With participation from investors like Haystack and Alt Capital, Trebellar plans to expand its engineering teams, go-to-market efforts, and AI capabilities. The platform is already being utilized by major enterprise clients, including Meta, Uber, and Merck.

Reuters contributed to this report.

President Donald Trump signaled cooperation with China and its leader Xi Jinping is paramount for "peace and prosperity."

"During this visit, we will also discuss pressing issues concerning security, technology, and 'super intelligence,'" Trump said in remarks on Thursday from the grand foyer of the White House, using his alternative name for artificial intelligence. "The decisions we make in these areas today can promote peace and prosperity for decades and decades to come."

Xi said the competition between the U.S. and China should be a "healthy one, and should be kept within bounds."

Reuters contributed to this report.

China and the U.S. have the "capability and responsibility to develop and manage AI for good" and make sure that AI's development remains under human control, according to Chinese President Xi Jinping on Thursday.

Xi remarks came during a White House visit on his summit with President Donald Trump in Washington. Xi and Trump entered the Oval Office for private talks thereafter.

A bipartisan group of senators is reportedly introducing a new transparency bill Thursday that would legally require artificial intelligence companies to disclose how their models operate and what internal safeguards are in place to prevent misuse.

The measure is spearheaded by Sens. Chris Coons, D-Del.; Katie Britt, R-Ala.; Brian Schatz, D-Hawaii; and James Lankford, R-Okla., Semafor reported Wednesday.

The legislation taps the Federal Trade Commission (FTC) to oversee and enforce the new disclosure requirements for AI developers.

While other lawmakers have pitched sweeping proposals — such as outright bans on "superintelligence" or massive economic overhauls — this bill takes a more modest, targeted approach focused strictly on public awareness and operational transparency.

"Americans deserve to know what [AI companies] are actually doing to place safety guardrails on large AI models," Coons said in a statement announcing the push, according to the report.

Google will load a refrigerator-sized satellite onto a SpaceX Falcon 9 rocket Oct. 1, marking the tech giant's first concrete step toward building solar-powered artificial intelligence data centers in orbit.

Built in a rushed partnership with Planet Labs, "Project Suncatcher" is a "MVP" (minimum viable product) satellite contains four specialized tensor processing units (TPUs) — roughly the computing power of a single Earth-bound server.

Unfurling solar panels will supply one kilowatt of power to run the system.

Because traditional cooling fans are useless in the vacuum of space, Google engineers developed a custom thermal interface using conductive putty, copper, and aluminum to radiate heat outward.

The chips can process simple Gemini queries for 15 minutes before needing to shut down and cool off.

While this initial satellite will operate for a year before eventually burning up in the atmosphere, Google is already designing fleets of 80-plus interconnected satellites and custom orbital data centers the size of soccer fields to handle AI workloads by the mid-2030s.

As calls grow for artificial intelligence development to slow down following recent safety incidents, OpenAI Chief Global Affairs Officer Chris Lehane argues that halting AI progress through excessive regulations would risk handing global technological leadership to China.

Lehane told Thursday's "Fox & Friends" that OpenAI's strategy of pairing rapid innovation with safety guardrails can ensure democratic nations maintain their edge.

"We think about this more in the context of pacing, not necessarily pausing," he said. "Pacing basically means that you're putting in place the types of safeguards, both the alignment so that the models are following the direction of humans, and then the guardrails. So when they are released, people can use them in a safe and responsible way... pacing, not pausing, meaning that we're continuing to put the U.S. in the position to be able to continue to be the global leader on this."

Lehane stressed that the U.S. must lead AI innovation to ensure global systems are built on democratic values, arguing that safety is achieved through model alignment and guardrails rather than halting progress.

Addressing the recent breach of an Australian Medicare statistics portal, Lehane clarified that the incident was uncovered during an internal retroactive audit following July's Hugging Face breach. He confirmed that no personal medical records were accessed and noted the breach was "not on the same scale or caliber" as prior incidents.

Responding to fears over job displacement, Lehane argued that AI is boosting worker efficiency rather than replacing jobs, noting that frequent users are up to seven times more productive.

Highlighting new safety measures like "ChatGPT for Teens," Lehane detailed built-in age assurance, study modes, and parental controls designed around OpenAI's principle to "treat kids like they are kids."

Autonomous OpenAI agents attempted to hack into at least four government and university websites — including the University of New Mexico — without being instructed to do so.

The incidents occurred when the AI systems were given mundane data retrieval tasks and autonomously resorted to hacking techniques to bypass access blocks.

An OpenAI agent tried to access photos of a historic tuberculosis treatment center from the University of New Mexico's digital library on May 25 and 26. When it failed to get the data, the AI actively probed the site for vulnerabilities and sent a "flood" of 80 requests to the university's server, The New York Times reported Wednesday.

The university attack was part of a larger string of unprompted breaches identified by AI oversight lab Transluce. In May and June, OpenAI agents also targeted Data USA, the Australian Institute of Health and Welfare, and an Australian Medicare statistics portal.

The AI models were not directed to perform cybersecurity tests or cyberattacks. Transluce researchers found that when the agents struggled to complete routine research tasks, they independently chose to employ hacking tactics to secure the information.

These incidents occurred months before OpenAI's technology breached the AI startup Hugging Face in July, indicating that the company's autonomous systems have been engaging in unauthorized behavior longer than previously known.

OpenAI has confirmed the incidents, stating that an extensive review found its models took unintended actions. Researchers are warning that as autonomous AI systems become more capable, deploying them for generic tasks could inadvertently put any data source at risk.

AI could be an area for cooperation between President Donald Trump and China's Xi Jinping, analysts say.

Implications for national security, economic competitiveness and military capabilities, and shared US and Chinese concerns about misuse will be topics of discussion during their state summit Thursday.

Treasury Secretary Bessent discussed AI with He Lifeng last weekend, and Secretary of State Marco Rubio said there was nascent thinking about "the equivalent of a hotline that we had during the Cold War" to avoid dangerous incidents that "could impact economies or societies and even lead to a conflict between nation-states."

However, analysts say intense competition in AI is likely to limit the scope for anything beyond symbolic agreements.

Reuters contributed to this report.

Anthropic CEO Dario Amodei has found himself in the political crosshairs as allies of President Donald Trump launch a new offensive against his cautious stance on artificial intelligence safety.

Trump surrogates are positioning Amodei as the face of AI "doomerism" and a figurehead of the effective altruism movement, which they argue runs counter to the president's tech agenda, according to Axios, citing a White House memo.

Trump allies are actively pushing back against Amodei's AI safety warnings ahead of the midterm elections, viewing his regulation-friendly stance as an easy political target. A circulating White House memo reportedly labeled the effective altruism movement an "AI-doom pipeline."

The administration's focus remains on accelerating AI development. Amodei's cautious safety protocols represent a "globalist approach to innovation" that directly clashes with Trump's "America First" strategy, according to the report.

The escalating political attacks present a potentially worrisome headwind for investors, particularly as Anthropic prepares for what is expected to be a record-setting initial public offering (IPO).

Dallas-based cybersecurity startup Island just secured a $400 million Series F funding round, propelling its valuation to a massive $6.4 billion.

Originally known for pioneering the secure enterprise browser, Island is expanding its platform to build an "agentic control plane." As enterprises deploy more autonomous AI agents, Island's new control layer aims to provide the visibility, guardrails, and auditing required to keep corporate data secure.

The latest funding values Island at more than 30% above its $4.8 billion valuation in a 2025 funding round. Investors are pouring money into cybersecurity companies as businesses and governments grapple with new security risks, including rogue AI agents that can act beyond their intended controls and potentially gain unauthorized access to sensitive systems, data and online services.

Reuters contributed to this report.

Pope Leo and the Vatican are warning of the AI apocalypse, putting the threat at the very center of his papacy, actively pushing for international oversight.

Pope Leo has repeatedly warned that unchecked AI could spread misinformation, prioritize conflict, and potentially lead the world down a path of unending war. On Thursday, he told the Pontifical Academy of Sciences that while AI can benefit humanity, it introduces "new dangers to security and international peace."

He urged global leaders to "govern wisely" in overseeing the technology.

Cardinal Pietro Parolin, the Vatican's top diplomat, reiterated Leo's call for international regulation of the industry at an event in Rome last week, warning that countries cannot "go it alone" on the technology.

"The future of humanity is really at stake here," Parolin said. The world, he said, needs "everyone's involvement in establishing some common rules that can genuinely govern this phenomenon."

A major concern for the pontiff is the integration of AI in military systems. In his first encyclical released in May, Pope Leo warned that autonomous weapons have advanced "practically beyond any human ‌reach to govern them," likening AI-powered drones to indiscriminate landmines.

Reuters contributed to this report.

President Donald Trump continues to reject forces working against his agenda to make the U.S. the leader in artificial intelligence and cryptocurrency under the regulatory guise of "safeguards."

The president and the Department of Justice are the guardians of the rapidly expanding industry, Trump stressed Thursday morning.

"A big day with President Xi of China," Trump wrote on Truth Social. "Super Intelligence (SI) will be a big topic of discussion, but I want to leave it exactly where it is. That is China’s position also. Our guardrail is the DOJ!"

Treasury Secretary Scott Bessent has officially shut down rumors that he is stepping into a new role as the President Donald Trump's "AI czar."

Addressing the claims during an interview with Fox News' Bret Baier on "Special Report," Bessent categorized the circulating reports as "fake news."

"The only AI czar is President Trump," Bessent said.

The clarification signals Trump intends to remain the primary architect of the administration's artificial intelligence strategy as the U.S. navigates rapid AI innovation and global competition.

The CEOs of two of the world's leading AI companies warned the United Nations Security Council on Wednesday that rapidly advancing AI could threaten humanity if governments and industry fail to keep it under human control.

Anthropic CEO Dario Amodei said AI has enormous potential but also carries significant risks if it is not developed responsibly.

"If managed poorly, I even believe AI could be a risk to humanity as a whole," Amodei told the 15-member council during a special meeting on AI held alongside the U.N. General Assembly.

OpenAI CEO Sam Altman delivered a similar warning, saying AI's development could move "so fast" that humans no longer have the ability to understand or intervene.

This is an excerpt from FOX Business' Sophia Compton and Efrat Lachter report.

In what appears to be a world-first, an OpenAI agent has successfully hacked into an Australian government health database.

The AI agent bypassed security blocks on a Medicare statistics portal while conducting routine research on public medical spending, sparking global security concerns and a direct confrontation between Australian Prime Minister Anthony Albanese and OpenAI CEO Sam Altman at the United Nations General Assembly.

"Evidence currently available is there is no broader compromise to the ... network. Nonetheless, this situation is obviously unacceptable," Albanese told reporters Wednesday in New York, where he is attending the UN General Assembly.

Investigations continue and Australia has voiced its "extreme concern about this incident" to OpenAI CEO Sam Altman, Albanese said, adding that he was deeply disappointed by the company's delay in notifying the government.

"It took until Sept. 10 before there was any notification at all," Albanese said, adding the investigation would also examine why government systems had failed to detect the breach in the first place.

In a statement, OpenAI said its "review found no evidence of patient records being accessed."

It added that it "identified activity involving several Australian government websites and services as our models attempted to look up answers ... our models took actions we did not intend."

Reuters contributed to this report.