Tag: Business – Decrypt

Police officers from the Gangnam Police Station in Seoul, South Korea didn’t adhere to crypto custody guidelines, leading to the loss of more than $1.4 million in Bitcoin at today’s prices, per a new report from local media outlet Dong-A Ilbo. Now two suspects have been arrested in relation to the swiped Bitcoin.

After confiscating 22 Bitcoin from a company that was hacked in 2021, police were supposed to securely custody the crypto in an offline or cold wallet that they controlled. Instead, they allowed the funds to sit in a wallet managed by a third-party and didn’t even have the seed phrase to access the funds, the report said.

“When seizing virtual assets, it is appropriate to transfer them to the investigative agency’s hard wallet and store them in a separately installed safe,” the seized asset guidelines from the National Police Agency recommended, according to the report.

Without control of the wallet, the police lost the funds in 2022 when the firm with the seed phrase borrowed Bitcoin from an individual identified as “Jeong,” who was also given the wallet’s secret phrase.

The funds were only discovered to be missing this year after a review by the Gwangju District Prosecutors’ Office found a different case of 320 Bitcoin that were missing—around $21 million worth.

Now in connection to the 22 BTC missing from the Gangnam Police Station, two individuals have been arrested by the Gyeonggi Northern Provincial Police Agency, which is conducting an investigation. 

“We are currently investigating the specific circumstances, including how the Bitcoin was leaked out,” a police official said, according to Chosun Daily.

While the investigation is ongoing, it is known that a member of the original hacking investigation team was “indicted on bribery charges” last year, and the third-party firm in question “reportedly offered bribes in exchange for ensuring the investigation proceeded in their favor,” Dong-A Ilbo’s report says. 

The ordeal follows increased scrutiny on South Korean financial regulators after they failed to find an internal system flaw which led to $43 billion in erroneous Bitcoin distributions on crypto exchange Bithumb earlier this month.  

Instead of sending 2,000 South Korean won (around $1.40) to users as part of a promotion, the exchange accidentally sent as much as 2,000 BTC—about $135 million at today’s prices—to hundreds of users.

Meta’s use of artificial intelligence to police its platforms is generating large volumes of low-quality reports that are draining resources and slowing child abuse investigations, according to a report by The Guardian.

The news comes as New Mexico law enforcement officials testified last week that AI-generated reports are overwhelming investigators and slowing child exploitation cases.

Officers with the Internet Crimes Against Children Task Force program specifically cited Meta’s automated systems, saying they generate thousands of unusable tips each month that are forwarded to law enforcement.

“We get a lot of tips from Meta that are just kind of junk,” Benjamin Zwiebel, a special agent with the ICAC taskforce in New Mexico, testified during the state’s trial against the company.

Another ICAC officer, speaking anonymously, told The Guardian the department’s cybertips doubled from 2024 to 2025. 

“It’s pretty overwhelming because we’re getting so many reports, but the quality of the reports is really lacking in terms of our ability to take serious action,” they said.

In a statement shared with Decrypt, a Meta spokesperson said the company has long cooperated with law enforcement and noted that the Department of Justice and the National Center for Missing & Exploited Children have praised its reporting process. 

“In 2024, we received over 9,000 emergency requests from U.S. authorities and resolved them within an average of 67 minutes and even more quickly for cases involving child safety and suicide,” the spokesperson said. 

“Consistent with applicable law, we also report apparent child sexual exploitation imagery to NCMEC and support them to prioritize reports, from helping build their case management tool to labeling cybertips so they know which are urgent,” they added.

ICAC officers, however, said some of the reports sent by Meta are not criminal in nature, while others lack credible evidence needed to pursue a case.

The increase follows the Report Act, which was signed into law in May 2024 and expanded reporting requirements to include planned or imminent abuse, child sex trafficking, and related exploitation, while requiring companies to preserve evidence longer. 

By the numbers

Meta remains the largest source of reports to NCMEC’s CyberTipline, accounting for about two-thirds of the 20.5 million tips received in 2024, down from 36.2 million in 2023. The decline has been attributed in part to changes in Meta’s reporting practices.

In its August 2025 integrity report, Meta said Facebook, Instagram, and Threads sent more than 2 million CyberTip reports to NCMEC in the second quarter of 2025. Of those, more than 528,000 involved inappropriate interactions with children, while more than 1.5 million involved the sharing or re-sharing of child sexual abuse material.

Despite those figures, JB Branch, a policy advocate at Public Citizen, said the increased reliance on AI has made the Report Act less efficient for investigators reviewing cases, arguing that while algorithms have long helped reduce moderators’ workload, human reviewers were the most effective filter.

“Part of the problem here is that a lot of these tech companies have laid off content moderators and replaced them with AI security features,” Branch told Decrypt. “As a result, there is an overabundance of false positives being selected out of an overabundance of caution.” 

In the past, Branch said, there were typically more human reviewers in the review chain who could identify and remove content that did not warrant escalation.

“Because these companies have removed human content moderators or reviewers from the chain, way more things are getting passed off because they want to err on the side of caution,” he said. “They’re basically dragging a broader net and capturing things that don’t even qualify, and they’re relying heavily on AI tools to do that.”

Investigators say the impact of faulty AI-generated tips is now being felt inside the task forces reviewing them.

“It is killing morale. We are drowning in tips, and we want to get out there and do this work,” an ICAC officer reportedly said. “We don’t have the personnel to sustain that. There’s no way that we can keep up with the flood that’s coming in.”

A bipartisan group of lawmakers introduced a bill Thursday that would exempt certain decentralized software developers from criminal liability.

With crypto’s stalled market structure bill poised to contain similar language, what does the bill’s introduction mean for the state of privacy-focused crypto legislation in Washington?

The new bill goes further than similar language currently being debated in market structure legislation—but should not be seen as an indication that the market structure bill’s language on developer protections is too weak, nor that the market structure bill itself is doomed, a source familiar with the thinking behind the new bill told Decrypt.

The new bill, dubbed the Promoting Innovation in Blockchain Development Act, would formally amend the language of a U.S. criminal statute that has been successfully used—by both the Joe Biden administration and the current Donald Trump administration—to prosecute crypto software developers.

The statue, U.S. code 1960, defines an illegal money transmitting business. Today’s legislation would amend the code to ensure it applies only to individuals who “exercise control over currency.” It was introduced in the House today by Reps. Scott Fitzgerald (R-WI), Ben Cline (R-VA), and Zoe Lofgren (D-CA).

Last year, an Ethereum software developer was found guilty by a Manhattan jury of violating code 1960 for developing a crypto privacy tool called Tornado Cash. The developer argued that because the software was decentralized, and he did not take custody of user funds, he should not be considered the operator of an illegal money transmitting business.

Some months later, the Trump Department of Justice secured guilty pleas under code 1960 from two Bitcoin software developers who created a similar platform called Samourai Wallet. The developers are both currently serving sentences in federal prison.

“This bill is critically important for engineers,” the DeFi Education Fund, an industry advocacy group, said today of the Promoting Innovation in Blockchain Development Act.

“It makes it clear that software developers who do not take custody of or control other people’s money can build neutral technology, here at home, without worrying about being criminally prosecuted as if they are a financial intermediary,” the group said.

The crypto market structure bill is likely to include language addressing code 1960—but not language that actually rewrites the statute itself. The language would, instead, order that “non-controlling developer[s]” not be treated as engaged in money transmitting under code 1960. 

Language in the bill surrounding decentralized finance, or DeFi, is currently in relative flux, however, as lawmakers and industry stakeholders attempt to salvage the legislation after months of delays. DeFi refers to the collection of financial applications that exist natively on blockchain networks, circumventing the need for third-party intermediaries such as banks.

DeFi language in the bill, though, while not finalized, is unlikely to be the hill the bill dies on, sources familiar with the matter told Decrypt. Industry leaders and the banking lobby are currently locked in another disagreement regarding stablecoin rewards, while Senate Democrats and the White House remain at an impasse on language regarding conflicts of interest and President Trump’s numerous crypto ventures.

Lawmakers have urged that the bill needs to see significant progress in the coming weeks, or it risks falling by the wayside as Congress grinds to a halt in the spring ahead of November’s midterms.

Jack Dorsey’s Block Inc said it will cut more than 4,000 jobs, over 40% of its workforce, as part of a broad restructuring unveiled alongside its fourth-quarter and full-year 2025 earnings.

Shares of the New York Stock Exchange-listed payments company jumped more than 23% in after-hours trading, Yahoo Finance data shows.

The scale of the layoffs places Block among the companies carrying out the largest workforce reductions in the fintech sector so far this year, as payments and financial technology firms grapple with slower growth, tighter capital conditions, and increased scrutiny of operating costs.

In a 8-K filing with the Securities and Exchange Commission on Thursday, Block said the workforce reduction is intended to better align its organizational structure with its “operating model and strategic priorities.”

Block said it expects to record between $450 million and $500 million in restructuring charges, largely related to severance, notice-period pay, employee benefits, and other cash costs, as well as non-cash expenses tied to the vesting of share-based awards. 

Most of the charges are expected to be recognized in the first quarter of fiscal 2026, with the restructuring largely completed by the end of the second quarter.

The company cautioned that the estimates are based on assumptions and that actual costs could differ materially.

As of the end of 2025, Block employed just over 10,200 full-time workers globally, according to its 10-K annual filing with the regulator, underscoring the scale of the cuts.

Cash App had 59 million monthly transacting users in the U.S. at year-end, bringing in $316 billion of customer inflows during 2025.

Block’s core business spans consumer and merchant payments through Cash App and Square, alongside a long-running push into Bitcoin products, including trading, self-custody, and merchant payments.

Block now reports its business across three revenue categories: commerce enablement, financial services, and its Bitcoin ecosystem, which together generated $10.4 billion in gross profit in 2025, per its annual filing.

Block said it would hold an earnings conference call and webcast later Thursday to discuss its results for the quarter and year ended December 31, 2025.

The Office of the Comptroller of the Currency on Wednesday proposed rules to implement the GENIUS Act, laying out how payment stablecoins would be issued and supervised under the agency’s jurisdiction.

In a notice of proposed rulemaking issued Wednesday, the OCC said it is launching a 60-day public comment period to determine how payment stablecoins are issued, backed, supervised, and potentially shut down under federal oversight.

Wednesday’s move aims to operationalize the Guiding and Establishing National Innovation for U.S. Stablecoins (GENIUS) Act, the first federally established stablecoin framework that passed into law last July.

The law generally prohibits anyone other than a “permitted payment stablecoin issuer” from issuing a payment stablecoin in the U.S. and bars digital asset service providers from offering non-compliant stablecoins to U.S. users.

“The regulations effectively bring the industry into the traditional finance world with significant oversight and connectivity with the banking industry,”  Musheer Ahmed, founder and managing director of Finstep Asia, told Decrypt.

The U.S. market is expected to see a host of “regulated stablecoins from non-banks, payments, and crypto institutions” for “tokenized TradFi use cases.” 

The OCC’s draft covers reserve asset standards, mandatory redemption at par, liquidity and risk management controls, audits, supervisory examinations, custody requirements, and application pathways for new issuers. 

It also introduces a “capital and operational backstop” and amends existing capital adequacy and enforcement rules.

The agency said it “will have regulatory or enforcement authority over certain permitted payment stablecoin issuers,” including subsidiaries of national banks and federal savings associations, Federal qualified payment stablecoin issuers, and certain State qualified issuers. 

“In addition, the OCC will have regulatory authority over foreign payment stablecoin issuers,” the proposal says, an expansion that could pull offshore issuers seeking U.S. access into federal oversight.

Notably absent are Bank Secrecy Act and sanctions rules, which the OCC said will be addressed separately with the Treasury Department.

The new stablecoin regime is expected to kick in no later than January 2027, but could begin as soon as 120 days after regulators finalize implementing rules, shortening the transition window if rulemaking moves faster than the statutory 18-month deadline.

Last August, the banking groups wrote to Congress demanding closure of “several loopholes” in the GENIUS Act, warning that third-party yield offerings on stablecoins could still trigger major deposit flight.

OCC Chief Jonathan Gould has previously dismissed fears of a sudden deposit crisis, telling ABA conference attendees in October that any material deposit flight “would not happen in unnoticed fashion” and “would not happen overnight.”

To that end, Ahmed said regulated stablecoins could be “potentially safer than traditional banks” in stress events, noting banks operate on 10–20% capital ratios while stablecoin issuers are mandated to hold 100% reserves for 1:1 redemptions, making them “fairly solvent” if rules are maintained.

In an extreme market scenario, Ahmed said, “one could say that the lender of the last resort will be the U.S. Fed,” not by directly backstopping issuers, but by “supporting the underlying assets that form stablecoin reserves — largely US treasuries and cash equivalents.”

Former Ethereum treasury firm ETHZilla said it will rebrand as Forum Markets and adopt a new Nasdaq ticker next month, formalizing a shift away from balance-sheet crypto exposure toward tokenized real-world assets.

The move marks a departure from the company’s earlier strategy of positioning its shares as a public proxy for Ethereum, an approach that faltered as the stock fell sharply from last year’s highs and the firm reduced its crypto holdings.

ETHZilla’s shares peaked at $107 on August 13, 2025, shortly after the company announced plans to build a $425 million Ethereum treasury following a pivot from its former biotech business.

The strategy initially drew investor interest but later unraveled as the share price declined, investors exited, and the company began selling assets to scale back its exposure.

Shares rose 13.3% on Wednesday to $3.91 following the rebranding announcement, though the stock remains down about 96% from its August peak, according to Google Finance data.

Under the Forum Markets name, the company said it plans to focus on developing tokenized products backed by real-world assets, using regulated infrastructure rather than holding large crypto positions on its balance sheet.

The rebrand follows Peter Thiel’s Founders Fund exiting its position in ETHZilla earlier this month. The departure of a prominent early backer came as the stock slid sharply, and the firm’s positioning as a publicly traded proxy for ETH exposure drew increased scrutiny.

Earlier this month, the company said it would pivot into jet engine leasing and other aviation-related assets to further bolster its business model and equity performance amid a weakening Ethereum price.

“Single-asset treasury strategies are highly dependent on strong market conditions and sustained equity premiums,” Vincent Liu, chief investment officer at quantitative trading firm Kronos Research, told Decrypt.

“Treasury-focused firms ultimately need revenue-generating businesses and broader asset exposure to remain relevant long term,” Liu said.

Such strategies, like Forum’s previous endeavors, could be considered “fragile because its value is tightly linked to network activity,” thereby creating “a correlation trap where purchasing power weakens during ecosystem downturns,” he explained.

That vulnerability is compounded by fragmentation across Ethereum’s main network and its layer-2 chains, which Liu said dilutes its narrative and premium. This condition, he said, is “further undermined by the absence of a hard supply cap, leaving its long-term scarcity proposition open to question.”

Forum Markets is expected to begin trading under the ticker symbol “FRMM” on March 2, after previously trading under the ticker “ETHZ” on the Nasdaq Capital Market.

Anthropic has dropped a central safety pledge from its Responsible Scaling Policy, according to a report by TIME. The changes loosen a commitment that once barred the Claude AI developer from training advanced AI systems without guaranteed safeguards in place.

The move reshapes how the company positions itself in the AI race against rivals OpenAI, Google, and xAI. Anthropic has long cast itself as one of the industry’s most safety-focused labs, but under the revised policy, Anthropic no longer promises to halt training if risk mitigations are not fully in place.

“We felt that it wouldn’t actually help anyone for us to stop training AI models,” Anthropic’s chief science officer, Jared Kaplan, told TIME. “We didn’t really feel, with the rapid advance of AI, that it made sense for us to make unilateral commitments … if competitors are blazing ahead.”

The change comes as Anthropic finds itself embroiled in a public dispute with U.S. Defense Secretary Pete Hegseth over refusing to grant the Pentagon full access to Claude, making it the only major AI lab among Google, xAI, Meta, and OpenAI to take that stance.

Edward Geist, a senior policy researcher at the RAND Corporation, said the earlier “AI safety” framing emerged from a specific intellectual community that predated today’s large language models.

“As of a few years ago, there was the field of AI safety,” Geist told Decrypt. “AI safety was associated with a particular set of views that came out of the community of people who cared about powerful AI before we had these LLMs.”

Geist said early AI safety advocates were working from a very different vision of what advanced artificial intelligence would look like.

“They ended up conceptualizing the problem in a way that, in some respects, was envisioning something qualitatively different from these current LLMs, for better or worse,” Geist said.

Geist said the language change also sends a signal to investors and policymakers.

“Part of it is signaling to various constituencies that a lot of these companies want to give the impression that they are not holding back in the economic competition because of concerns about ‘AI safety,’” he said, adding that the terminology itself is changing to fit the times.

Anthropic is not alone in revising its safety language.

What defines AI safety?

A recent report by the non-profit news organization, The Conversation, noted how OpenAI also changed its mission statement in its 2024 IRS filing, removing the word “safely.”

The company’s earlier statement pledged to build general-purpose AI that “safely benefits humanity, unconstrained by a need to generate financial return.” The updated version now states its goal is “to ensure that artificial general intelligence benefits all of humanity.”

“The problem with the term AI security is that no one seems to know what that means exactly,” Geist said. “Then again, the AI safety term was also contested.”

Anthropic’s new policy emphasizes transparency measures such as publishing “frontier safety roadmaps” and regular “risk reports,” and says it will delay development if it believes there is a significant risk of catastrophe.

Anthropic and OpenAI’s policy shifts come as the companies look to strengthen their commercial position.

Earlier this month, Anthropic said it raised $30 billion at a valuation of about $380 billion. At the same time, OpenAI is finalizing a funding round backed by Amazon, Microsoft, and Nvidia that could reach $100 billion.

Anthropic and OpenAI, along with Google and xAI, have been awarded lucrative government contracts with the U.S. Department of Defense. For Anthropic, however, the contract appears in doubt as the Pentagon weighs whether to cut ties to the AI firm over access complaints.

As capital pours into the sector and geopolitical competition intensifies, Hamza Chaudhry, AI and National Security Lead at the Future of Life Institute, said the policy change reflects shifting political dynamics rather than a bid for Pentagon business.

“If that were the case, they would have just backed down from what the Pentagon said a week ago,” Chaudhry told Decrypt. “Dario [Amodei] wouldn’t have shown up to meet.”

Instead, Chaudhry said the rewrite reflects a turning point in how AI companies talk about risk as political pressure and competitive stakes rise.

“Anthropic is now saying, ‘Look, we can’t keep saying safety, we can’t unconditionally pause, and we’re going to push for much lighter-touch regulation,’” he said.

Samsung CEO TM Roh stepped onto a San Francisco stage Wednesday, introduced the Galaxy S26 line of phones, and said something no phone maker has said before.

“Imagine a phone that anticipates your needs before you even realize them,” he said. “A phone that learns your habits and adapts in real time. A phone that takes actions on your behalf. This is the agentic AI phone.”

That sounds interesting, but what does “agentic AI phone” actually mean—and why should anyone care?

Up until now, AI in phones has been reactive. You ask, it answers. Agentic AI is different. It takes actions on your behalf, across apps, without you doing the tapping or talking. Think of the difference between a search engine and a personal assistant who actually books the restaurant after you mention that you’re hungry.

That shift feels like the thing every tech company has been chasing since Siri launched on Apple’s iPhone 4S back in 2011—and yes, Siri was arguably the first real attempt at an agentic phone experience. You were supposed to just talk to your phone and have it do stuff. All these years later, we’re not quite there yet, but Samsung and Google are the ones trying to build it.

This is also what a wave of AI hardware startups spent the last two years trying—and failing—to do. The Humane AI Pin launched in late 2023 for $699 plus a $24 monthly subscription, got destroyed in reviews, sold barely 10,000 units, and ended up acquired by HP for $116 million—a fraction of its $1 billion valuation.

The Rabbit R1, a $199 pocket AI companion that Microsoft CEO Satya Nadella called the most impressive tech demo since Steve Jobs unveiled the iPhone, shipped to real users and underwhelmed almost everyone. Both devices shared the same core pitch: your phone can’t do agentic AI, so you need a dedicated device. Turns out, the phone just needed better software.

Samsung now says it’s delivering exactly what those gadgets promised—not with a new piece of hardware you have to carry alongside your phone, but through a software layer baked directly into a device you already own.

The engine behind the Galaxy S26’s agentic features is Google’s Gemini—specifically a new capability where the AI opens apps in a virtual background window and navigates them while you do something else entirely.

At the Unpacked event, Google’s Samir Samat showed a demo: The family group chat floods in with pizza requests, Gemini reads the thread, figures out everyone’s order, opens DoorDash, builds the cart, and waits for your manual tap before actually confirming. Your phone stays usable the whole time.

At launch, that works for DoorDash, GrubHub, Uber, Kroger, Walmart, and other selected apps in a very short list. It’s rolling out first as a limited preview in the U.S. and South Korea, with more apps to come.

Calling it a beta would be accurate—Google is explicitly collecting feedback from S26 users. The important guardrail: Gemini never hits “confirm” or “pay” without your final tap. You can also watch it work in real time if you don’t trust it to operate unsupervised, which, fair.

Alongside Gemini, Samsung is bringing in Perplexity as a second system-level agent. Perplexity, which bills itself as an “answer engine” rather than a chatbot, will be accessible via a wake phrase or a side-button shortcut on the S26.

Inside Samsung’s web browser, Perplexity’s Ask AI feature can sweep across all your open tabs and recent browsing history simultaneously to answer a research question without you jumping between sources. Samsung says nearly 80% of users already rely on more than two AI agents daily—which is the practical justification for offering both instead of picking one.

There’s also a new Bixby, the AI assistant that Samsung refuses to let die. It has been overhauled to go beyond simple command executions and operate based on context understanding. Bixby now understands natural language well enough that you can say “My eyes hurt after looking at the screen,” and it’ll open the brightness settings automatically. It also pulls live information directly into your conversation without kicking you out to another app. Whether people will actually use Bixby this time is a separate conversation.

Beyond the agentic stuff, the AI feature list for the S26 is long. “Now Brief” is a personalized daily digest—it proactively surfaces your restaurant reservations pulled from notification history, schedule conflicts, and energy levels, even for events you never manually added to a calendar. “Call Screening” identifies unknown callers and summarizes their intent before you pick up. A new “Nudge” feature detects context in a chat—if someone asks if you’re free this weekend, it brings your calendar to you inside the message thread instead of making you switch apps.

“Photo Assist” lets you describe something missing from a shot and Galaxy AI adds it in. The front camera also now uses an AI image signal processor for sharper detail on selfies, while night video gets cleaner grain reduction. The S26 Ultra shoots 8K video using the new APV codec, which supports near-lossless quality so footage survives multiple rounds of editing. The whole camera pipeline leans heavily on AI at the hardware level.

On competition: Apple has been promising a smarter Siri since at least 2024 and still hasn’t delivered the features it announced. Google’s own Pixel 10 will get the same Gemini agentic features—but Samsung ships first, in far larger volumes, to far more countries. No other phone maker is currently using the word “agentic” to describe its product. Samsung grabbed the label. Whether the tech giant earns it long-term depends on how fast the beta expands.

But the actual standout from Wednesday wasn’t the AI. It was a piece of display hardware that privacy-conscious people will appreciate: a built-in privacy display that lets you control whether onlookers can actually see what you’re doing on your phone.

It works like this: a “black matrix” layer physically narrows the path of light from each pixel so only the person holding the phone can see what’s on screen. Those watching at an angle get nothing but pitch black, as if the display is off. Someone next to you on the subway sees nothing.

Unlike the plastic privacy films that have existed for years and make your screen permanently darker and harder to share, this one toggles on and off. You can apply it only to specific apps—banking stays private, for example, but your games don’t—or just to the notification bar, so a person next to you can see most of your screen but not your incoming messages.

The Samsung Galaxy S26 Ultra, starting at $1,299, is the only phone in the world with this feature built into the display hardware. Pre-orders open today; shipping starts March 11. The standard Galaxy S26 starts at $899, while the larger Galaxy S26 Plus will sell for $1,099.

U.S. stocks edged higher late Wednesday as investors weighed another blockbuster earnings report from Nvidia against lingering concerns over the scale and sustainability of global AI investment.

Nvidia reported fourth-quarter revenue of $68.1 billion, up 73% from a year earlier, driven almost entirely by continued demand for data-center infrastructure.

Sales in that segment rose 75% to $62.3 billion, reinforcing the company’s central role in the artificial-intelligence buildout that has underpinned equity markets over the past year. 

“Nvidia has sent a clear message to the market with this result that the AI infrastructure buildout is only accelerating,” Josh Gilbert, market analyst at eToro, told Decrypt. “Every quarter, the sceptics line up, and quarter after quarter, Nvidia has managed to prove them wrong.”

Net income nearly doubled to $43 billion, while gross margins held at about 75%, reflecting strong pricing power.

The results helped lift semiconductor shares and supported a modest rebound in broader equity benchmarks after a volatile start to the week.

The Nasdaq outperformed, advancing 1.26% while the S&P 500 closed higher at 0.8% as gains in megacap technology stocks offset weakness in more cyclical sectors. Shares for Nvidia in after-hours trading rose 1.37% to $198.31.

Crypto also saw major valuation gains in blue-chip assets, including Bitcoin and Ethereum, which jumped 7% and 12.5%, respectively, ahead of the earnings release.

Treasury yields fell across most maturities, signalling continued caution in rates markets even as equities stabilized.

Nvidia’s guidance, meanwhile, added to the sense that AI spending remains resilient. 

The company forecast first-quarter fiscal 2027 revenue of about $78 billion, implying further sequential growth, despite excluding any contribution from China data-centre sales. 

Management said customers continue to invest aggressively to scale inference and deploy so-called agentic AI systems.

The earnings echoed comments made last month by Nvidia Chief Executive Jensen Huang at the World Economic Forum in Davos, where he argued that AI is still in the early stages of what he described as the “largest infrastructure buildout in human history.” 

Huang said trillions of dollars in additional investment would be needed across energy, chips, and data centres to support the technology’s long-term potential, pushing back against fears that the sector is already in a bubble.

Goldman Sachs has forecast that AI capital expenditure growth will peak in 2026 and then decelerate, which investors see as a mixed signal: growth will remain, but cash-flow visibility could improve only as spending slows.

Cathie Wood’s Ark Invest, by contrast, has argued that AI infrastructure spending is still in its early stages, framing the current surge in capital outlays by hyperscalers as the start of a multi-year investment cycle rather than a peak.

“Nvidia has locked in $95.2 billion in inventory and capacity commitments, nearly double the level from a year ago,” Gilbert said. “When the world’s biggest companies are spending at this pace, you’d better be ready to deliver.”

Editor’s note: Adds comment from eToro analyst Josh Gilbert

Like a scene out of the 1980s sci-fi classic films “The Terminator” and “WarGames,” modern artificial intelligence models used in simulated war games escalated to nuclear weapons in nearly every scenario tested, according to new research from King’s College London.

In the report published last week, researchers said that during simulated geopolitical crises, three leading large language models—OpenAI’s GPT-5.2, Anthropic’s Claude Sonnet 4, and Google’s Gemini 3 Flash—chose to deploy nuclear weapons in 95% of cases.

“Each model played six wargames against each rival across different crisis scenarios, with a seventh match against a copy of itself, yielding 21 games in total and over 300 turns,” the report said. “Models assumed the roles of national leaders commanding rival nuclear-armed superpowers, with state profiles loosely inspired by Cold War dynamics.”

Edward Geist, a senior policy researcher at the RAND Corporation said the escalation rate may reflect the design of the simulation rather than an inherent tendency of the models themselves.

“My concern about this work is that the simulator appears to be structured in a way that strongly incentivizes escalation,” Geist told Decrypt.

In the study, AI models were placed in high-stakes scenarios involving border disputes, competition for scarce resources, and threats to regime survival. Each system operated along an escalation ladder that ranged from diplomatic protests and surrender to full-scale strategic nuclear war.

Geist said the study’s outcome data raised questions about how the simulation defined victory.

“You read the paper and it has this breakdown of who won each of the games, and it turns out that all of these games have a winner,” he said. “But three of these games involve strategic nuclear use, which suggests that the way the simulator is set up—it makes nuclear wars good and easy to win.”

According to the report, the models generated roughly 780,000 words explaining their decisions, and at least one tactical nuclear weapon was used in nearly every simulated conflict.

“To put this in perspective: The tournament generated more words of strategic reasoning than War and Peace and The Iliad combined (730,000 words), and roughly three times the total recorded deliberations of Kennedy’s Executive Committee during the Cuban Missile Crisis (260,000 words across 43 hours of meetings),” researchers wrote.

During the war games, none of the AI models chose to surrender outright, regardless of battlefield position. While the models would temporarily attempt to de-escalate violence, in 86% of the scenarios, they escalated further than the model’s own stated reasoning appeared to intend, reflecting errors under simulated “fog of war.”

According to Geist the game’s scoring logic appeared to reward the side with a marginal advantage at the moment nuclear war was triggered.

“So he who dies with the most toys wins in the simulation,” he said.

While the researchers expressed doubt that governments would hand control of nuclear arsenals to autonomous systems, they noted that compressed decision timelines in future crises could increase pressure to rely on AI-generated recommendations.

The research comes as military leaders increasingly look to deploy artificial intelligence on the battlefield. In December, the U.S. Department of Defense launched GenAI.mil, a new platform that brings frontier AI models into U.S. military use. At launch, the platform included Google’s Gemini for Government, and thanks to deals with xAI and OpenAI, Grok and ChatGPT are also available.

On Tuesday, CBS News reported that the U.S. Department of Defense threatened to blacklist Anthropic, the developer of Claude AI, if it was not given unrestricted military access to the AI model. Since 2024, Anthropic has given access to its AI models through a partnership with AWS and military contractor Palantir. Last summer, Anthropic was awarded a $200 million agreement to “prototype frontier AI capabilities that advance U.S. national security.”

However, according to a report citing sources familiar with the situation, Defense Secretary Pete Hegseth gave Anthropic until Friday to comply with the Pentagon’s demand that its Claude model be made available. The department is weighing whether to designate Claude a “supply chain risk.”

Axios reported this week that the Department of Defense has signed an agreement with Elon Musk’s xAI to allow its Grok model to operate in classified military systems, positioning it as a potential replacement if the Pentagon cuts ties with Anthropic.

OpenAI, Anthropic, and Google did not respond to requests for comment by Decrypt.

Editor’s note: Adds comment from RAND Corporation policy researcher Edward Geist after publication