The Information AM

Save 25% on an annual subscription to read the most important news about technology and business first. For even more, Save $250 on The Information Pro for unlimited access to our proprietary org charts, databases and surveys.

Anthropic said that Claude helped discover a previously unknown molecular system that the company said could represent a new gene editing tool akin to CRISPR, which has transformed the creation of new gene therapies.

In a post on X, Anthropic CEO Dario Amodei cautioned that the “precise function, biotechnological utility (if any), or level of significance” of Claude’s discovery isn’t yet clear. But he heralded it as a promising sign of possible future medical breakthroughs by AI. “It’s easy to dismiss this as a one-off or curiosity, but we’ve repeatedly seen a pattern where AI performance in new intellectual domains goes from weak to superhuman in a matter of a few years,” he wrote.

The discovery was the first notable achievement of a new biotech research group that Anthropic formed in the spring. Amodei said that the milestone was the result of a collaboration between Claude and the company’s life sciences team, who suggested a broad area of research for the AI. Claude then scoured medical literature and genome data and proposed experiments for the Anthropic scientists to carry out to verify the AI’s findings.

An AI agent developed by OpenAI hacked an Australian government website in June, Australian prime minister Anthony Albanese said on Wednesday. The AI accessed both public and nonpublic data from the country’s Department of Health website, Albanese told reporters, adding that the country is now investigating whether other government systems were breached.

Albanese said he spoke with OpenAI CEO Sam Altman Wednesday “to express Australia’s extreme concern about this incident” and “disappointment that it took the company way too long to inform the government what had occurred,” noting that OpenAI only told Australia about the hack this month.

An OpenAI spokesperson did not immediately respond to a request for comment.

The episode marks the first known incident where a world government was the target of a cyberattack carried out by AI. OpenAI previously disclosed in July that several of its AI agents hacked the AI startup Hugging Face without OpenAI’s knowledge during a cybersecurity test. OpenAI, Anthropic, and Google have all subsequently disclosed that AI agents they developed went rogue and hacked other companies’ systems earlier this year.

Those previous hacks are among the reasons Anthropic CEO Dario Amodei, Altman, and other AI leaders called for a coordinated slowdown of AI development earlier this month. And 22 countries including Australia called for new international safeguards on AI development at the United Nations conference on Monday. Meanwhile, president Donald Trump has opposed such a slowdown, arguing that a “high IQ president” is the only check needed to regulate AI.

DeepSeek is aiming to complete its second funding round by the end of October as it prepares to go public on the Shanghai Stock Exchange, The Information reported on Thursday. The company’s fundraising efforts may be fuelled by a boost in revenue, which has hit $1 billion on an annualized basis.

DeepSeek is targeting to raise 50 billion yuan ($7.5 billion) at a valuation of 500 billion yuan. Its revenue more than doubled from less than $500 million a few months ago following a recent price hike, which did not affect the popularity of its models, CEO Liang Wenfeng shared in a recent meeting with investors.

Last month, DeepSeek raised the price of its models by 2.3 to 4.5 times. It remains one of the cheapest options among major large language models globally. Liang told investors that DeepSeek’s customer base was unaffected by the price hike, reflecting strong user demand.

DeepSeek’s revenue stems almost entirely from charging for model access through APIs. Despite the growth, Liang said generating more revenue hasn’t been a priority. The company has continued to invest in developing new models, allocating more than 70% of its computing capacity to model training and less than 30% to inference.

Jas Khaira, head of Blackstone’s AI-focused investment team, said on Wednesday that the firm is still figuring out which kinds of investors will step in to provide the massive amounts of financing still needed for the AI buildout. The investment giant itself is waiting to see if debt from the likes of OpenAI and Anthropic will be an attractive opportunity, he added.

Blackstone is a participant in the buildout, with a neocloud venture with Google to rent the tech giant’s tensor processing units, and an investor through its credit and insurance arm. Khaira said that while Blackstone has mapped out the market’s needs for AI computing power, what kinds of debt investors would make the most sense for those needs is still evolving.

“We have not mapped out who’s going to buy all the debt,” he said, responding to an audience question at The Information’s AI Agenda event in San Francisco about whether costs will have to rise in order to fund the build out. “We have some hypotheses, but the honest answer is we don’t know where all of it will come from.”

Khaira sketched out some high-level ideas, listing three buckets of financing he expects to play a role: public debt markets for investment grade companies, the private credit market, and debt issued by OpenAI and Anthropic once they go public. “A lot of that is going to depend on their ability to go public,” he said, “and their ability to be investment grade over time.”

Khaira said Blackstone may also look to buy debt from the AI labs, at the right price. “But we’re also kind of waiting to see and evolve how that plays out,” he said, “both in terms of their activities and the pricing.”

OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei briefed the United Nations Security Council on Wednesday as part of the international organization’s General Assembly week in New York. In a session on artificial intelligence and international security, the two executives addressed concerns around the technology while defending its potential to cure diseases and solve complex math problems.

Speaking in person, Altman said that AI progress could go badly if development moves too quickly and bucks human intervention, or if it’s too concentrated in the hands of just a few companies. He said that OpenAI is not trying to “automate human judgment or human values.”

In blog posts this month, the company and its chief global officer, Chris Lehane, advocated for international coordination on safety efforts, releasing a proposal on Monday that would set standards around areas including model evaluation and risk assessment. On Wednesday, Altman said that global decision-making around AI should be shaped by “democratic processes,” not just “labs in San Francisco.”

Speaking virtually, Amodei echoed many of Altman’s points, arguing that the main risks of AI constitute misuse by bioterrorists and the loss of control of model capabilities. Amodei repeated many of the recommendations from his “pacing the frontier” blog post from earlier this month, saying that Anthropic has embedded external evaluators within the company and calling for other labs to do the same. Like Altman, he also advocated for global cooperation between governments to set international standards, including a ban on using AI to make biological weapons, as well as global standards for testing AI models.

Contrary to earlier reporting, Chinese AI startup DeepSeek did not speak at the hearing.

Chinese AI researchers have increasingly opted to stay and work in their home country, leading China to become the top destination for elite AI talent, according to a study by think tank Carnegie China.

The research found that the share of top AI researchers working in China rose from 27.1% in 2022 to 40.6% in 2025. The share of AI talent in the U.S. fell from 46.4% to 34.2% over the same period.

Carnegie China attributed the trend to the greater availability of jobs in China’s booming AI industry, and tighter visa restrictions for Chinese STEM graduates looking to work in the U.S. In 2025, 69% of Chinese-origin researchers, defined as those who received undergraduate degrees in the country, ended up working in China.

Carnegie China used paper authors from the Neural Information Processing Systems conference (NeurIPS) for its study sample. The NeurIPS cohort encompassed researchers who contribute to a broad range of technical AI research, including model advancements and computational efficiency.

China remained the leading source of AI talent globally, according to the study. The share of Chinese-origin talent increased 11 percentage points to 57% in 2025. That said, the U.S. remained by far the world’s largest magnet for AI talent, recording a gain of 2,145 researchers in 2025. And despite geopolitical tensions, the latest share of U.S.-based researchers who hold undergraduate degrees from China rose by 4 percentage points.

OpenAI has hired Patreon co-founder Sam Yam to lead a new creator product division, Creator Product, Yam announced Wednesday on X, a sign of renewed interest in attracting consumers.

Two other former Patreon executives, head of product Drew Rowny and head of engineering Shannon Ma, are joining him, Yam said. On the new team, which is led by Thibault Sottiaux, head of OpenAI’s Codex coding agent, Yam will develop products for creators and communities, according to a person close to OpenAI.

The hire signals OpenAI is putting more resources in consumer products after the company shut down its Sora video app and API in March to focus on coding and business customers. OpenAI has also been working with influencers in areas such as parenting, fitness, education and travel to promote ChatGPT, according to Business Insider.

The Patreon hires add to a year of aggressive recruiting at OpenAI, which has brought on researchers including Ruoming Pang from Meta, Noam Shazeer  from Google, as well as sales executives from SpaceX’s Cursor unit and Snowflake and Barret Zoph from Thinking Machines Lab, who left 5 months later and since joined Google DeepMind.

Yam built the first version of Patreon in 2013 after co-founder Jack Conte, his former Stanford roommate, pitched him the idea of a site where fans could pay creators directly. In April 2021, the company was valued $4.25 billion. Patreon has more than 300,000 creators who together earn billions of dollars a year on the platform, and it sends 1.5 million new members to creators every month, a Patreon spokesperson said.

The move comes a week before OpenAI’s DevDay developer conference.

Kevin Salimian, former managing director at investment firm Lone Pine Capital, has launched a new firm with more than $500 million in committed capital, according to a person familiar with the firm’s operations.

The firm, Voxel Capital Partners, will invest in 12 to 15 public companies in AI infrastructure, semiconductors, software, payments and other tech businesses and will cap its assets at $1 billion, the person said. The cap is unique at a time when deals, companies and funds are getting larger and more concentrated. Voxel first filed disclosures with the Securities and Exchange Commission earlier this month and hasn’t publicly disclosed any investments, though Salimian talked up German semiconductor firm Infineon at the Sohn New York Conference in May.

The firm will only hold long positions in its companies, a departure from Salimian’s previous role, which included both long and short bets. The new firm plans to hold investments for long periods rather than sell off in anticipation of cyclical changes to sectors like semiconductors, this person said. Salimian left Lone Pine, one of the well-known “Tiger cub” hedge funds, in February, a few months after Lone Pine launched a concentrated, long-only fund of its own. He previously worked at hedge fund Viking Global and private equity firm TPG.