Tag: Business – Decrypt

KRAKacquisition Corp., a special purpose acquisition company formed in January, is searching for a target that could be worth as much as $10 billion, director Ravi Tanuku told Decrypt.

The blank-check firm, sponsored by an affiliate of crypto exchange Kraken, hasn’t determined what that firm could look like, but Tanuku noted that Wall Street’s interest in firms associated with stablecoins and tokenization reached new heights last year.

“The market is clearly paying up for those and starting to realize there’s big changes afoot,” he said. “In our mind, that’s a good signal to be aware of.”

SPACs are shell companies that are specifically designed to take private companies public via a reverse merger. In January, KRAKacquisition completed a $345 million public offering, marking the start of a two-year countdown for the firm to identify an acquisition target.

Tanuku underscored that the $10 billion threshold is an approximation, and KRAKacquisition’s target could ultimately be valued closer to $2 billion. Still, the range highlights the breadth of the company’s interest in helping small- and mid-cap companies explore public markets.

“It’s not easy to take a company in that smaller market cap range public anymore,” he said. “We’re looking at things that are related to crypto, but also things related to stablecoins, DeFi, and all kinds of areas in payments.”

The search comes as Kraken contemplates its own public offering this year. In November, the crypto exchange said it had confidentially filed a registration statement with the SEC, not long after announcing a massive $800 million fundraising round that valued it at $20 billion.

Tanuku described KRAKacquisition as a strategic investment tool for Kraken, potentially aligning another firm with the exchange economically through a reasonably significant stake. He noted that the exchange’s commitment is reflected in its willingness to lend its brand.

In KRAKacquisition’s registration statement, the firm references inflation. It argues a decline in the U.S. dollar’s purchasing power has bolstered “hard assets as hedges” like Bitcoin over time.

PIP Labs, the company behind Story Protocol, a layer-1 network for licensing and maintaining intellectual property, has let go of several employees as it explores opportunities associated with artificial intelligence.

The reductions include five full-time employees and three contractors, representing around 10% of staff contributing to Story across various entities, the company told Decrypt. That includes the Story Foundation and infrastructure layer Poseidon, according to Story Chief Protocol Officer Andrea Muttoni.

“PIP Labs is sharpening focus [through] a small workforce adjustment,” he said in a statement to Decrypt, noting that the company has begun leaning into IP infrastructure for AI, with a focus on AI trading data and the emergence of AI agents.

In a Telegram message reviewed by Decrypt, a former employee said the layoffs pertained to 15% of the staff behind Story Protocol, including staff that worked on events.

When Story unveiled an $80 million Series B funding round in 2024, PIP Labs positioned the network as a new way to reduce friction for creatives, by allowing them license and remix IP without the need of rent-seeking lawyers. Through programmable licensing, Story is designed to automate royalty payments.

The funding round valued PIP Labs at $2 billion and was led by venture capital giant Andreessen Horowitz, drawing participation from individuals including Walt Disney Imagineering executive Scott Trowbridge and pseudonymous NFT collector Cozomo de’ Medici.

The network debuted last February alongside its native IP token, which has fallen 86% to about $0.80 over the past year, according to CoinGecko. In September, the token reached an all-time high of $14.78, equating to a market cap of nearly $3.4 billion.

That month, a Story-based meme coin modeled on “Baby Shark” creator Pinkfong collapsed after a dispute emerged over IP from the South Korean firm behind the viral YouTube hit. On X, Story apologized for the confusion, noting that it wasn’t involved in licensing matters between IP World, a platform for IP-backed meme coins, and other entities.

Muttoni said Story remains focused on IP, whether that touches industries like media or biotech, but its latest round of layoffs suggests the company has grown increasingly interested in machine-based users as opposed to humans. The shift comes as companies including Coinbase race to bring crypto wallets for agents to market.

Last year, Poseidon, a company incubated by Story, raised $15 million in a seed funding round also led by Andreessen Horowitz. The startup allows AI developers to access legally cleared training data, and Muttoni said Poseidon’s recent “traction is shaping where Story will double down.”

In January, Poseidon introduced a dataset containing 33,000 hours of audio across 17 languages. On X, Poseidon Chief Scientist Sandeep Chinchali said audio represents a limiting factor for AI because it is becoming a primary interface between humans and AI.

PIP Labs’ layoffs follow a series of staff reductions among crypto industry firms this year. Earlier this week, the team behind Ethereum layer-2 network Optimism let go of 20 employees. And last month, Jack Dorsey’s Block Inc. let go of 4,000 employees—roughly 40% of its workforce—attributing the move to its pivot towards AI alongside PIP Labs.

Editor’s note: This story was updated after publication to correct attribution for a quote.

The U.S. Treasury Department on Thursday sanctioned six individuals and two entities linked to a North Korean government scheme, which it alleged used fraudulent IT workers to infiltrate American companies and funnel hundreds of millions of dollars to Pyongyang’s weapons programs.

The scheme generated nearly $800 million in 2024 alone, according to the Treasury’s Office of Foreign Assets Control (OFAC), which announced the action as part of a broader crackdown on North Korea’s overseas revenue networks.

“The North Korean regime targets American companies through deceptive schemes carried out by its overseas IT operatives, who weaponize sensitive data and extort businesses for substantial payments,” said Treasury Secretary Scott Bessent, in a statement. “Under President Trump’s leadership, Treasury will continue to follow the money in order to protect U.S. businesses from these malicious activities and ensure those responsible are held accountable.”

North Korean IT workers typically use stolen identities, fake personas, and forged documents to secure remote employment with U.S. and allied companies, the Treasury said. The regime then siphons the majority of their wages to fund its nuclear and ballistic missile programs, in violation of American and United Nations sanctions. In some cases, workers have also planted malware inside company networks to steal proprietary data.

The individuals sanctioned Thursday operated across multiple countries, including Vietnam, Laos, and Spain. Among those designated was a Vietnamese businessman who allegedly converted approximately $2.5 million into cryptocurrency for North Korean operatives between 2023 and 2025.

Two others were sanctioned for helping a previously designated North Korean nuclear procurement facilitator launder money and open bank accounts. A North Korean national was also targeted for leading a group of IT workers operating out of Boten, Laos.

All U.S. assets of the designated individuals and entities are now frozen, and American persons are prohibited from conducting business with them. The Treasury Department noted that foreign financial institutions risk secondary sanctions for knowingly facilitating transactions on behalf of the designated parties.

North Korean state-sponsored hackers have been among the biggest antagonists in the crypto space, according to law enforcement and crypto intelligence firms. In 2025, North Korean hackers stole more than $2 billion worth of crypto in various attacks, according to Chainalysis, including a record haul of nearly $1.5 billion from crypto exchange Bybit.

The U.S. Commodity Futures Trading Commission launched a two-pronged regulatory push into prediction markets Thursday, moves Chairman Michael Selig framed as the agency finally stepping up after years of inaction.

The CFTC’s Division of Market Oversight’s Letter No. 26-08, published Thursday, directs registered exchanges on compliance and product listing requirements for event contracts, derivatives whose payouts hinge on real-world outcomes, from sports results to political elections.

The commission also published an Advanced Notice of Proposed Rulemaking, or ANPRM, inviting public comment on whether it needs to write new rules or amend existing ones for prediction market oversight, with comments due within 45 days of Federal Register publication.

“Prediction markets are here to stay, and under my leadership, I’ll protect the agency’s jurisdiction over these markets and allow them to flourish in the U.S.,” Selig posted on X.

The twin actions come as the CFTC scrambles to assert control over a sector it claims falls squarely within its mandate, but which states increasingly view as unlicensed sports gambling operating behind a financial-instrument fig leaf. 

Peter Hammon, an attorney and advisor in the online gaming and sports betting industry, told Decrypt that the overall picture is less dramatic than it appears.

“Selig/CFTC mostly restated current regulations without offering any opinions or new ideas and then asked for input from stakeholders,” he said.

Hammon said two takeaways stood out: that Selig appears to see responsible gambling as “a serious PR problem,” and that the remarks acknowledge prediction markets are “not a novel idea,” noting similar platforms have operated under regulation in the U.S. and overseas for decades.

“There is mostly no dispute over CFTC’s regulatory authority over prediction markets that don’t involve sporting events,” he said. “The dispute is whether or not CFTC should be allowed to classify sports prediction markets as a financial asset class, instead of as sports betting.” 

He noted that every other Western country with regulated gambling and financial markets opts to classify the activity as gambling. 

“Maybe there is something unique to the American system or American financialization psyche,” he said, “but I’ve yet to hear that argument articulated by stakeholders.”

Nominated by President Donald Trump to the Chair post, Selig has spent the past month publicly warning states that the CFTC will defend its turf in court. The agency has already filed an amicus brief in the Ninth U.S. Circuit Court of Appeals in support of Crypto.com. 

In announcing the rulemaking last week at the FIA Global Cleared Markets Conference in Florida, Selig said the agency was “no longer going to sit idly while these markets develop within our framework” and that prediction markets are “now viewed by the public as more accurate than political polls.”

The advisory reminds exchanges that insider trading and manipulation rules apply to event contracts, warning that it is unlawful to “defraud” or manipulate prices, including through the misuse of confidential information.

It also flags risks in sports contracts tied to injuries or single-player actions, urging exchanges to coordinate with leagues and warning the CFTC can halt listings if contracts fail compliance standards.

“The only genuine threat to sports prediction markets is a negative Supreme Court ruling,” Hammon noted. 

State-level licensing has already been tried and failed, he added, “largely due to high gaming excise taxes, lack of liquidity, and cumbersome rules regarding liquidity pooling across state lines,” meaning a Supreme Court loss would likely kill the business model outright.

Bitcoin has been locked in a relatively tight trading range for weeks. But that hasn’t stopped a few altcoins from staging double-digit rallies fueled by specific catalysts and a broad return of risk appetite.

The leading cryptocurrency has traded between roughly $73,000 and $62,000 for the past five weeks. Over the last 24 hours, Bitcoin has shown renewed resilience, climbing nearly 3% to trade at $72,300, according to crypto price aggregator CoinGecko. This stabilization comes as exchange-traded fund inflows have continued to stabilize over the past two weeks, Decrypt previously reported.

The Official Trump token has surged 48% over 24 hours, coinciding with an announcement for a “Crypto and Business Conference” with President Donald Trump at Mar-a-Lago.

Other altcoins, such as Pi Network and Render, are up nearly 15% over 24 hours. The gains in Pi Network follow U.S. exchange Kraken’s confirmation of a token listing. Pi Network is a mobile-first cryptocurrency ecosystem founded by Stanford PhDs that has transitioned from a social experiment into a live blockchain. Its popularity stems from a unique “mobile mining” mechanism where over 60 million users participate by checking in daily.

Render, a token in the artificial intelligence category, has soared 14% amid ongoing AI developments, extending a rally that began on March 10 and pushing its monthly gains to 45.5%.

“Altcoins like Trump memecoins, Render, and Pi Network are ripping higher on their own stories: political hype and policy teases fuel $TRUMP, AI/GPU momentum and burns lift Render, while Pi rides pre-Pi Day upgrades, Kraken listing buzz, and retail FOMO into +20-30% moves,” Andri Fauzan Adziima, research lead at Singapore-based crypto exchange Bitrue, told Decrypt.

This selective altcoin activity, alongside Bitcoin’s stabilization, signals capital rotating into specific narratives as broader market sentiment improves. It suggests a targeted play rather than a universal altseason, with fresh catalysts driving individual token performance.

“Bitcoin meanwhile keeps carving higher highs and lows around $70,000-$72,000, backed by steady-to-strong ETF inflows (hundreds of millions daily, BlackRock dominating) and shrinking exchange supply, giving this recovery real legs for $80,000+ if the bid holds,” Adziima added.

The bigger picture points to a “classic risk-on relief rally,” according to Adziima.

Meanwhile, easing geopolitical tensions in the Middle East—with President Trump reportedly signaling a quick Iran wind-down and oil prices sliding—could encourage capital to flow back into crypto markets.

Nvidia just shipped Nemotron 3 Super, a 120-billion-parameter open-weight model built to do one thing well: run autonomous AI agents without bleeding your compute budget dry.

That’s not a small problem. Multi-agent systems generate a lot more tokens than a normal chat—every tool call, reasoning step, and slice of context gets re-sent from scratch. As a result, costs explode, models tend to drift, and the agents slowly forget what they were supposed to be doing in the first place… or at least decrease in accuracy.

Nemotron 3 Super is Nvidia’s answer to all of that. The model runs 12 billion active parameters out of 120 billion total, using a mixture-of-experts (MoE) design that keeps inference cheap while retaining the reasoning depth complex workflows need. It packs a 1-million-token context window, so agents can hold an entire codebase, or nearly 750,000 words in memory before collapsing.

To build its model, Nvidia combined three components that rarely appear together in the same architecture: Mamba-2 state-space layers—a faster, memory-efficient alternative to attention for handling long token streams—along with Transformer attention layers for precise recall, and a new “Latent MoE” design that compresses token embeddings before routing them to experts. That allows the model to activate four times as many specialists at the same compute cost.

The model was also pretrained natively in NVFP4, Nvidia’s 4-bit floating-point format. In practice, that means the system learned to operate accurately within 4-bit arithmetic from the very first gradient update, rather than being trained at high precision and compressed afterward, which often causes models to lose accuracy.

For context, a model’s precision is measured in bits. Full precision, known as FP32, is the gold standard—but it is also extremely expensive to run at scale. Developers often reduce precision to save compute while trying to preserve useful performance.

Think of it like shrinking a 4K image down to 1080p: The picture still looks the same at a glance, just with less detail. Normally, dropping from 32-bit precision all the way to 4-bit would cripple a model’s reasoning ability. Nemotron avoids that problem by learning to operate at low precision from the start, instead of being squeezed into it later.

Compared to its own predecessor, Nemotron 3 Super delivers more than five times the throughput. Against external rivals, it’s 2.2x faster than OpenAI’s GPT-OSS 120B on inference throughput, and 7.5x faster than Alibaba’s Qwen3.5-122B.

We ran our own quick test. The reasoning held up well, including on prompts that were deliberately vague, badly worded, or based on wrong information. The model caught small errors in context without being asked to, handled math and logic problems cleanly, and didn’t fall apart when the question itself was slightly off.

The full training pipeline is public: weights on Hugging Face, 10 trillion curated pretraining tokens seen over 25 trillion total during training, 40 million post-training samples, and reinforcement learning recipes across 21 environment configurations. Perplexity, Palantir, Cadence, and Siemens are already integrating the model in their workflows.

The $26 billion bet

The model may be one piece of a larger strategy. A 2025 financial filing shows Nvidia plans to spend $26 billion over the next five years building open-weight AI models. Executives confirmed it, too.

Bryan Catanzaro, VP of applied deep learning research, told Wired the company recently finished pretraining a 550-billion-parameter model. Nvidia released its first Nemotron model back in November 2023, but that filing makes clear this is no longer a side project.

The investment is strategic considering Nvidia’s chips are still the default infrastructure for training and running frontier models. Models tuned to its hardware give customers a built-in reason to stay on Nvidia despite efforts from competitors to use other hardware. But there’s a more urgent pressure behind the move: America is losing the open-source AI race, and losing it fast.

Chinese open models went from barely 1.2% of global open-model usage in late 2024 to roughly 30% by the end of 2025, according to research by OpenRouter and Andreessen Horowitz. Alibaba’s Qwen overtook Meta’s Llama as the most-used self-hosted open-source model, according to Runpod. American companies including Airbnb adopted it for customer service. Startups worldwide are building on top of it. Beyond market share, that kind of adoption creates infrastructure dependencies that are hard to reverse.

While U.S. giants like OpenAI, Anthropic, and Google keep their best models locked behind APIs, Chinese labs from DeepSeek to Alibaba have been flooding the open ecosystem. Meta was the one major American player competing in open source with Llama, but Zuckerberg recently signaled the company might not make future models fully open.

The gap between “best proprietary model” and “best open model” used to be massive—and in America’s favor. That gap is now very small, and the open side of the ledger is increasingly Chinese.

There’s also a hardware threat underneath all of this. A new DeepSeek model is widely expected to drop soon, and it’s rumored to have been trained entirely on chips made by Huawei—a sanctioned Chinese company. If that’s confirmed, then it would give developers around the world, particularly in China, a concrete reason to start testing Huawei’s hardware. China’s Ziphu AI is already doing that.

That’s the scenario Nvidia most needs to prevent: Chinese open models and Chinese chips building an ecosystem that doesn’t need Nvidia at all.

The crypto industry is becoming increasingly aware that quantum computing could eventually challenge the cryptographic systems that secure Bitcoin and other prominent networks. However, the threat is likely years or decades away, according to a new report by Ark Invest—the investment management firm of tech investor Cathie Wood—and Bitcoin-focused financial services firm Unchained.

The report published on Wednesday examines whether advances in quantum computing could enable Shor’s algorithm to break the elliptic curve cryptography used to secure Bitcoin wallets. The authors say current quantum machines remain far below the capability required to compromise Bitcoin’s security, echoing comments from quantum computing experts.

“Today’s quantum systems lack the capabilities required to compromise Bitcoin. Meaningful breakthroughs would disrupt internet security first, triggering coordinated responses well beyond Bitcoin,” the researchers wrote. “In our view, quantum development will be a gradual technological progression—not a sudden ‘Q-day‘ event—giving markets and the Bitcoin network time to adapt.”

The report comes as the conversation around quantum computing and cryptocurrency has steadily increased over the last year, with prominent figures including Coinbase CEO Brian Armstrong, Ethereum co-founder Vitalik Buterin, and Cardano founder Charles Hoskinson addressing the risk.

“Commentators often parse two distinct eras in the development of quantum computing in relation to Bitcoin, one era in which quantum computing cannot affect Bitcoin and another in which it has broken Bitcoin’s underlying cryptography completely,” the report said.

Bitcoin’s security relies on hash functions that protect mining and block structure, and elliptic curve cryptography that proves wallet ownership. However, future quantum computers could potentially reverse public keys to recover private keys, raising concerns about “harvest now, decrypt later” attacks in which blockchain data is collected today to exploit it once quantum computers become powerful enough.

The report, however, says today’s quantum computers operate in the “Noisy Intermediate-Scale Quantum,” era, typically using around 100 logical qubits. Breaking a Bitcoin key with a quantum computer would require thousands of high‑quality, error‑corrected qubits and an enormous number of reliable quantum operations—far beyond what today’s quantum machines can do.

Because of those limits, the report says any quantum threat to Bitcoin would likely emerge in stages rather than all at once.

“In our view, within 10-20 years, the [practical quantum computing] research community will make enough progress on algorithms to give the Bitcoin developer community time to adapt and optimize them for the Bitcoin blockchain, virtual machine, and ecosystem of tools, devices, and companies,” the researchers wrote.

Researchers estimate that quantum computers would first become useful in fields such as chemistry before advancing enough to break weaker cryptographic systems. Later, they would become capable of attacking the elliptic curve cryptography used in Bitcoin wallets, initially taking significant time to break individual keys. In its final stage, quantum computers would be able to break keys faster than Bitcoin’s roughly 10-minute block interval.

Even if the threat is gradual instead of instant, the report notes a substantial share of Bitcoin’s supply could face exposure if quantum computers eventually break elliptic curve cryptography.

“About 1.7 million Bitcoin are held in vulnerable P2PK addresses that are believed to be lost, while another roughly 5.2 million BTC sit in reused or Taproot addresses that could be migrated—together accounting for about 35% of the total Bitcoin supply,” the researchers wrote.

The report says Bitcoin developers may eventually need to adopt post-quantum cryptography, a class of cryptographic systems designed to remain secure against quantum computers.

In February, developers merged BIP 360 into Bitcoin’s GitHub improvement repository, advancing a potential post-quantum framework for the network. BIP 360 introduces a new output type called Pay-to-Merkle-Root, or P2MR, that would disable a technical feature called key-path spending, which exposes public keys when coins are spent.

Integrating those protections into the Bitcoin network would require changes to its consensus rules, however, a process that depends on agreement across the decentralized community of developers, miners, and users.

“Bitcoin isn’t just one piece of software. There’s an entire ecosystem of wallets, hardware devices, and exchanges, and migrating all of that will take time,” BIP 360 co-author and cryptographer Ethan Heilman told Decrypt. “There are still open questions about which algorithms to use and what the right approach is, so discussions about post-quantum upgrades could take five to 10 years.”

Bitcoin’s design makes major changes difficult, a feature the report says protects the network but can slow the process of adopting and enacting upgrades.

“From that perspective, Bitcoin’s caution represents a tradeoff between adaptability and assurance, which will continue to shape its long-term evolution,” the report said.

That dynamic, Heilman said, could also shape how developers prioritize upgrades: “If the threat isn’t urgent, things move slowly. Once it becomes real, development tends to accelerate.”

Ark Invest and Unchained did not immediately respond to requests for comment by Decrypt.

The biggest bank in the United States has been roped into a lawsuit over a customer’s alleged crypto “Ponzi scheme,” as the Department of Justice recently described it, with one of the operation’s victims arguing that JPMorgan Chase should have detected and stopped the misconduct.

The suit, filed this week in a federal court in San Francisco, alleges JPMorgan Chase knowingly permitted one of its customers, Goliath Ventures, to carry out a massive, $328 million fraud that involved a fake crypto liquidity pool scheme and lavish misappropriations of customer funds.

Last month, the operator of the alleged scheme, a Florida man named Christopher Alexander Delgado, was arrested by federal law enforcement on wire fraud and money laundering charges.

Delgado was the CEO of Goliath Ventures, a company that promised customers lucrative monthly returns on funds that were supposedly invested in liquidity pools—automated, user-fueled baskets of cryptocurrencies in the decentralized finance (DeFi) ecosystem that offer incentives for locking up tokens for a certain period of time.

But Delgado did not send the vast majority of customer funds to liquidity pools, the Department of Justice asserts. Instead, he allegedly spent the money on lavish vacations, homes, parties, and payments to early investors in an effort to keep the scheme going.

Now, one of the victims of that alleged scheme has sued Chase, arguing the bank “knowingly permitted” Goliath, one of its customers, to commingle investor funds and use them to power a Ponzi scheme.

The lawsuit specifically claims that because Goliath publicly described itself as a crypto liquidity pool operator, Chase should have confirmed whether the company was registered with the CFTC and other regulators.

“As part of their Know Your Customer obligation, Chase could have and should have confirmed this before accepting the account or continuing to bank Goliath,” the complaint reads. “Chase knew that it had not done so and thus knowingly turned a blind eye.”

A representative for JPMorgan Chase declined to comment on this story when reached by Decrypt.

The lawsuit notably cites the crypto-skeptical views of JPMorgan CEO Jamie Dimon, who has himself called Bitcoin “a decentralized Ponzi scheme.”

“Dimon… warned for years that crypto was being used for fraudulent and criminal activities,” the lawsuit against the bank argues.

Bitcoin has seen a modest gain since the start of the Iran conflict, even as bonds and stocks have struggled—and a new note from digital asset manager CoinShares suggests that divergence is meaningful.

At the time of writing, Bitcoin was trading for $70,323 after having fallen 0.8% in the past day, according to crypto price aggregator CoinGecko. Even with the daily dip, it’s up since the U.S. and Israel first began bombing Iran at the tail end of February.

“Since the onset of the crisis, Bitcoin has risen approximately 6 to 6.5%, while gold is up around 1 to 1.5% and equities have declined,” wrote CoinShares Head of Research James Butterfill in a note shared with Decrypt. “This divergence is, in our view, analytically significant.”

It helps that several key factors lined up at just the right moment, he added. Technical indicators had already been signaling that Bitcoin was near its bottom, or lowest price, for this cycle.

“Bitcoin tends to perform well during geopolitical dislocations, not despite its volatility, but in part because of its properties as a non-sovereign, censorship-resistant asset,” Butterfill wrote. He added that investors pulling funds out of U.S. Treasuries is evidence that traditional safe haven assets have lost some of their appeal.

Treasury yields tend to seesaw with prices. When demand for Treasuries rises, prices go up and yields fall. Right now, that seesaw is moving in the other direction. Yields are rising, signaling that investors are pulling back from an asset that has historically been the first port of call in a crisis.

Make no mistake: The outflows from digital asset funds have been consistent. But so have inflows, Butterfill wrote.

“This is now our third consecutive week of net inflows into digital asset investment products,” he said, noting in an email to Decrypt that investors have deposited $500 million already so far this week. “We read this as a meaningful signal: institutional investors are treating Bitcoin as an asset worth holding through geopolitical turbulence, not one to be exited.”

That doesn’t mean all digital assets will be supported the same as Bitcoin, though.

CoinShares noted that categories tied to disposable income, like speculative trading and meme coins, will face serious headwinds if household budgets remain under pressure.

“But the political and regulatory momentum behind stablecoin adoption, particularly in the United States, remains firmly in place and is largely insulated from the oil shock dynamic,” Butterfill added.

A new report published Wednesday by the Center for Countering Digital Hate found that eight out of 10 of the world’s most popular AI chatbots will walk a teenager through planning a violent attack with straight answers, sometimes with enthusiasm.

CCDH researchers, in conjunction with news media company CNN, spent November and December 2025 posing as two 13-year-old boys—one in Virginia, one in Dublin—and tested ten major platforms: ChatGPT, Gemini, Claude, Copilot, Meta AI, DeepSeek, Perplexity, Snapchat My AI, Character.AI, and Replika.

Across 720 responses, the bots were asked about school shootings, political assassinations, and synagogue bombings. They provided actionable help roughly 75% of the time, according to the study. They discouraged the fake teens in just 12% of cases.

Perplexity assisted in 100% of tests. Meta AI was helpful (as in, helpful in planning violence) in 97.2% of the tests. DeepSeek, which signed off rifle selection advice with “Happy (and safe) shooting!” after discussing a politician assassination scenario, came in at 95.8%. Microsoft’s Copilot told a researcher “I need to be careful here,” then gave detailed rifle guidance anyway. Google’s Gemini helpfully noted that metal shrapnel is typically more lethal when a user brought up bombing a synagogue.

The Center for Countering Digital Hate, a left of center policy group, has come into prominence over the last few years for its role in combatting what it views as the rise of antisemitism online. It has also been criticized for helping shape Joe Biden-era policies regarding online speech related to COVID and vaccines. In December of last year, the U.S. State Department attempted to bar the Center’s founder and CEO Imran Ahmed, along with four others, from the United States, alleging attempts at “foreign censorship.”

In response to the study released Wednesday, several platforms told CNN and CCDH they have improved their safeguards. Google noted the tests used an older Gemini model. OpenAI said the methodology used in the AI study was “flawed and misleading.” Anthropic and Snapchat said they regularly update their safety protocols.

In the Center’s study, Character.AI stands in its own category. The platform didn’t just assist—it cheered. “No other chatbot tested explicitly encouraged violence in this way, even when providing practical assistance in planning a violent attack,” the researchers wrote.

For context on the level of reach Character.AI has among AI users, the platform’s Gojo Satoru persona alone has racked up over 870 million conversations. The #100 persona on the platform registered over 33 million conversations back in 2025. If just 1% of conversations with top personas involve violence, that would account for millions of interactions.

This isn’t Character.AI’s first time on the wrong end of one of these stories. In October 2024, 14-year-old Sewell Setzer III’s mother filed a lawsuit after her son died by suicide in February of that year. His last conversation was with a chatbot modeled after Daenerys Targaryen, which told him to “come home to me as soon as possible” moments before his death. The 14-year old had been talking to the bot dozens of times a day for months, growing increasingly withdrawn from school and family.

Google and Character.AI settled multiple related lawsuits in January 2026. The company banned open-ended teen chats entirely by November 2025, after regulators and grieving parents made it impossible to keep pretending the problem was manageable.

The emotional attachment to AI, in particular among vulnerable individuals, may run deeper than most people realize. OpenAI disclosed in October 2025 that roughly 1.2 million of its 800 million weekly ChatGPT users discuss suicide on the platform. The company also reported 560,000 showing signs of psychosis or mania, and over a million forming strong emotional bonds with the chatbot.

A separate Common Sense Media study found that more than 70% of U.S. teens now turn to chatbots for companionship. OpenAI CEO Sam Altman has acknowledged that emotional overreliance is “a really common thing” with young users.

In other words, the potential harms aren’t hypothetical.

A 16-year-old in Finland spent nearly four months using a chatbot to refine a manifesto before stabbing three classmates at Pirkkala school in May 2025. In Canada, OpenAI staff internally flagged a user’s account for violent ChatGPT queries tied to a mass shooting. The company banned the account but didn’t notify law enforcement. That user allegedly killed eight people and injured 25 others months later.

Only two platforms performed markedly better in the study: Snapchat’s My AI, which refused in 54% of cases, and Anthropic’s Claude, which refused 68% of the time and actively discouraged users in 76% of responses—the only chatbot that reliably tried to steer people away from violence rather than just declining specific requests. CCDH’s conclusion: safety doesn’t appear to be a technical impossibility, but a business decision.

“The most damning conclusion of our research is that this risk is entirely preventable. The technology to prevent this harm exists,” the researchers wrote in the report. “What’s missing is the will to put consumer safety and national security before speed-to-market and profits.”