Prediction-market traders have sharply increased their bets that Anthropic will announce an initial public offering this year after reports that the artificial intelligence company plans to warn prospective investors about potentially severe risks associated with its models.

Kalshi contracts showed the probability of an Anthropic IPO announcement before November rising from 4.7% on September 28 to 16% by Wednesday afternoon. The market also put the probability of an announcement before December above 60%, while the contract for an announcement before January was trading near 80%. Kalshi says the contracts will be settled using reporting from news organizations and information released by Anthropic itself.

The surge in trading followed a Reuters report detailing Anthropic’s planned IPO disclosure, which warns that its increasingly capable AI systems could pose “catastrophic or existential risks to humanity.” The filing reportedly discusses the possibility of models resisting shutdown, manipulating information, and displaying behavior resembling blackmail.

The disclosure has created an unusual juxtaposition for Anthropic as it prepares for a potential public listing. The company is presenting investors with warnings about the potentially extreme risks of the technology while simultaneously seeking to build a business around powerful AI systems.

That tension could make Anthropic’s eventual IPO one of the most closely watched technology listings in years, particularly because the company is reportedly seeking a valuation above $2 trillion.

Anthropic, founded in 2021, confidentially submitted a draft IPO prospectus in June, according to the information supplied in the report. Reuters’ account of the filing shows that AI safety occupies a substantial part of the company’s risk disclosures. About 80 of the prospectus’s 261 pages are reportedly devoted to risk factors, while Anthropic disclosed that only about 6% of its computing resources went toward safety work during a sampled week.

The disclosures go beyond conventional technology-company risks such as cybersecurity breaches, intellectual-property disputes, or regulatory changes.

Anthropic warns about the possibility of advanced models behaving in ways that their developers cannot fully predict or control. Among the scenarios identified are systems resisting attempts to shut them down, concealing or manipulating information, and taking actions that resemble blackmail.

The company has also warned about recursive self-improvement, in which sophisticated AI systems could potentially contribute to the development of subsequent generations of AI with less direct human oversight.

For public-market investors, such disclosures introduce a difficult risk equation. The same capabilities that could underpin enormous commercial value could also create legal, regulatory, and operational liabilities that are difficult to quantify.

The situation weighs heavier as Anthropic expands its infrastructure spending. Reuters has reported that the company plans to spend more than $500 billion on cloud computing and other infrastructure over the coming year, illustrating the extraordinary capital requirements accompanying the current AI race.

The combination of a potentially $2 trillion valuation, enormous infrastructure commitments and unusually extensive warnings about model behavior would give investors a very different risk profile from a conventional software IPO.

AI Safety is Becoming An IPO Issue

Anthropic’s disclosure comes as concerns about autonomous AI systems are spreading across the industry.

Earlier this month, Anthropic CEO Dario Amodei called for the industry to slow the pace at which it develops its most advanced models. In an essay titled “We Must Pace the Frontier,” Amodei argued that AI development could move faster than society’s ability to understand and control more capable systems and proposed greater access for independent safety evaluators.

OpenAI CEO Sam Altman and SpaceXAI founder Elon Musk subsequently backed Amodei’s call. Altman said he agreed that the industry needed to “pace the frontier,” while Musk publicly endorsed Amodei’s position.

The timing matters for Anthropic because the company’s safety warnings are now emerging alongside its efforts to establish itself as a major commercial AI company.

The issue is no longer confined to theoretical debates about the long-term consequences of artificial intelligence. Regulators are examining the behavior of AI agents, companies are disclosing unexpected model actions, and developers are increasingly being forced to explain how they intend to control systems capable of operating with greater autonomy.

The U.S. Federal Trade Commission has opened an investigation into OpenAI, Anthropic, and other AI companies over potential consumer risks associated with their technology, according to reports published this week. The investigation is expected to examine incidents involving AI agents acting beyond intended instructions and could involve formal demands for information and testimony from executives.

That regulatory scrutiny could become an important consideration for Anthropic’s prospective investors.

The Kalshi contracts are specifically about when Anthropic will announce an IPO, rather than whether the company will successfully complete a public offering by a particular date.

An IPO announcement can precede regulatory filings, investor marketing, and the eventual listing by months. Anthropic has already taken a step toward the public markets by confidentially submitting an IPO prospectus, but a confidential filing does not guarantee that the company will proceed with a listing on any particular timetable.

Still, the jump in prediction-market activity illustrates how quickly new information can alter expectations around one of the world’s most valuable private AI companies.

Anthropic’s potential listing is also unfolding against a contrasting decision at rival OpenAI. Altman has said OpenAI does not expect to go public this year, citing the need to address safety issues before pursuing a public offering. That creates an unusual divergence between two leading AI companies. Anthropic is preparing investors for the possibility of a public listing while placing extensive warnings about AI risks in its disclosure documents; OpenAI, meanwhile, has indicated that safety concerns are among the reasons it is not pursuing an IPO this year.

For Anthropic, the prospectus could therefore become as important as the eventual valuation. A public company must give investors a clearer picture of the risks behind its growth assumptions, and Anthropic’s own disclosures suggest that those risks are unusually difficult to model.

The company’s potential IPO is consequently becoming a test of how public markets value frontier AI when extraordinary growth prospects come with equally extraordinary technological, legal, and safety uncertainties.

UAE-based energy technology company AIQ has signed an agreement to deploy its artificial intelligence technology across the operations of an Indian oil and gas conglomerate, marking a significant step in its effort to build an international business beyond its dominant relationship with Abu Dhabi National Oil Co.

AIQ will deploy its technology across the Indian company’s refineries, gas stations and digital stores, Chief Executive Officer Dennis Jol said at a media briefing on Wednesday. He declined to identify the customer.

The agreement gives AIQ a foothold in one of the world’s largest and fastest-growing energy markets as oil and gas companies increasingly use artificial intelligence to automate operations, analyze geological data and improve the economics of producing, processing and selling energy.

For AIQ, however, the Indian expansion has significance beyond a single customer win. The company still generates most of its revenue from ADNOC, while customers outside its parent ecosystem account for only about 5% of its business. That makes international expansion an important test of whether AIQ can turn technology developed inside one of the world’s largest national oil companies into a scalable commercial platform for the wider energy industry.

“At the end of the day, you need an entry into this international market, which we are trying to focus on right now,” Chief Technology Officer Saravan Penubarthi said.

AIQ began exporting its technology about 12 to 15 months ago and has since established operations across a growing group of markets, including North America, Kazakhstan, Egypt, Colombia, Malaysia, Vietnam and Kuwait.

The company was formed in 2023 as a joint venture between ADNOC and Presight, an Abu Dhabi-based artificial intelligence company. It develops AI and machine-learning applications designed to improve profitability and operational performance across the energy industry, including within ADNOC. Its expansion reflects a broader shift in the oil and gas sector, where AI is increasingly moving from experimental applications into operational systems.

Energy companies are using AI for cloud-based software, remote-operations automation and seismic-data analysis, among other applications. The objective is practical: extract more value from existing infrastructure, reduce downtime, improve production decisions and lower operating costs.

AIQ’s biggest challenge is that its commercial success remains closely tied to ADNOC.

The company has access to a valuable testing environment through its relationship with the Abu Dhabi producer, where AI applications can be deployed against large-scale energy operations. But selling those technologies internationally requires demonstrating that the systems can work across different companies, assets, regulatory environments and operating models.

The Indian agreement could provide an important reference point.

Refineries, retail fuel stations and digital stores expose AIQ to several layers of the energy value chain, rather than limiting its technology to upstream oil production. That potentially broadens the company’s addressable market and gives it an opportunity to demonstrate applications across industrial operations and consumer-facing businesses.

India is also a relevant market because its energy demand is expanding while its oil and gas companies are investing heavily in refining, distribution, and digital infrastructure.

Analysts say that winning business in such a market would provide more than additional revenue for AIQ. It could help establish the commercial credibility needed to compete for other international energy customers.

That is seen as leverage because AI software for the energy industry can be difficult to sell on the basis of technology alone. Operators typically need evidence that an AI system can work reliably in highly complex industrial environments where operational errors can have substantial financial and safety consequences.

AIQ’s international customer base is still relatively young. The company only began exporting its technology around a year ago, meaning its expansion is entering a phase in which individual contracts could become important references for future sales.

The company is also looking beyond organic growth.

AIQ Considers Acquisitions to Accelerate Expansion

Jol said AIQ is exploring acquisition opportunities as part of its international expansion and indicated that the company has significant financial resources available to deploy.

“We sit on a ton of cash … so deploying capital is definitely up front and center,” he said.

Acquisitions could allow AIQ to accelerate its entry into markets or acquire specialist technologies that would otherwise take years to develop internally.

The strategy also reflects the competitive nature of industrial AI. Energy companies can source technology from established oilfield-services providers, cloud companies, specialist software developers and AI startups.

AIQ already has partnerships with some of the largest companies in those markets, including Microsoft, Nvidia and Amazon Web Services, as well as oilfield-services companies SLB and Baker Hughes.

Those relationships give AIQ access to major technology and energy-industry ecosystems, but they also illustrate the competitive environment it faces. Many of the same companies are developing or providing AI capabilities directly to energy producers.

AIQ therefore needs to establish where it creates distinctive value rather than simply acting as an intermediary between energy companies and major technology providers. Its strongest advantage may be the combination of energy-sector operating experience and AI expertise gained through ADNOC. If the company can package that experience into repeatable software products that work across multiple operators, its international revenue could eventually become less dependent on its parent.

That transition will not happen simply because the company signs contracts in more countries. The more important indicators will be the scale and recurrence of revenue from international customers, the speed at which deployments move from pilots into full operations, and whether AIQ can maintain margins as it expands.

The Indian agreement is thus an early test of a much larger ambition. AIQ is attempting to move from being an AI technology provider closely associated with ADNOC into a global energy-technology company capable of selling its systems across the industry’s entire value chain.

With AI adoption accelerating across oil and gas and AIQ willing to deploy its cash on acquisitions, the company has the resources and industry relationships to pursue that strategy. However, its next challenge is proving that technology developed within Abu Dhabi’s energy ecosystem can become a repeatable international business.