Anthropic’s move to suspend access to its latest frontier AI models—after a U.S. directive tied to national security—has reignited debate over how concentrated control over advanced AI can translate into sudden access restrictions for users worldwide. Grayscale says the episode underscores the “centralized control” risks of frontier AI and may bolster demand for decentralized alternatives such as Bittensor.

In a note published Monday, Grayscale head of research Zach Pandl linked the U.S. order to the decision to cut access to Anthropic’s Fable 5 and Mythos 5 models for foreign nationals, and later to disable access for all users to comply with the directive. Pandl argued that investors increasingly want ways to access AI capabilities without relying on a single institution’s permissions.

Key takeaways

• Grayscale says Anthropic’s access suspension highlights the risk that centralized frontier AI can be curtailed quickly by governments.
• After the shutdown, Grayscale pointed to a surge in demand for decentralized AI networks, citing Bittensor’s TAO token strength.
• The U.S. order required Anthropic to suspend access for foreign nationals over national security concerns, prompting broader compliance measures.
• Bittensor is positioned by Grayscale as an “open, global, decentralized” approach to AI access.

U.S. directive forces Anthropic to pull access

According to the reporting referenced by Cointelegraph, the U.S. government directed Anthropic on Friday to suspend access to its models for foreign nationals, citing national security concerns. Anthropic then disabled access to Fable 5 and Mythos 5 not only for the targeted group, but for all users, to comply with the order.

Grayscale’s Pandl framed this as a stark example of how quickly centralized systems can change who is allowed to use the latest AI capabilities. In his Monday note, he said the situation “drives home the need for decentralized alternatives,” arguing that frontier AI access can be shaped by decisions made outside the open technical ecosystem.

Grayscale ties the access shock to momentum in decentralized AI

Pandl also pointed to market behavior in the immediate aftermath of Anthropic’s cut-off. He noted that in the roughly 12 hours after access was reduced, Bittensor’s TAO token rose by 30%, reaching a three-week high of $283 on Monday.

The note positions that move as evidence that users and investors are actively looking for alternatives when access to prominent centralized models is disrupted. CoinGecko data referenced in the original piece indicates TAO’s outperformance relative to the broader crypto market over the prior week.

While token price changes do not prove causality, the timing described by Grayscale suggests that decentralized AI networks may be perceived as more resilient when frontier model availability is constrained.

Bittensor as an “AI for everyone” network, not a permissioned gate

Grayscale’s research director argued that Bittensor offers a different approach to AI infrastructure: an “alternative vision for AI based on decentralized principles,” intended to deliver access to AI resources through an open, global, decentralized network.

Pandl summarized the concept by comparing it to “Bitcoin for AI,” emphasizing the idea that access to capabilities should be governed by open protocols rather than by a single lab’s authorization policies. In his view, as AI improves, AI access increasingly functions like an economic resource—meaning the rules determining who can use it, and under what conditions, become a central issue for both governments and market participants.

Industry voices: centralized AI limits are becoming more visible

Beyond Grayscale, other commentators highlighted broader implications. Colton Malkerson, co-founder of EdgeRunner AI, described the incident as a “breaking point” for corporate data independence, telling Cointelegraph in a note that companies are “renting” intelligence from big labs—an arrangement he said can become worse when access is withdrawn abruptly.

In his analogy, he framed it like a tenant whose landlord can cancel a lease at any time while also viewing the tenant’s property. The point was not about a specific model’s quality, but about the structural dependency that arises when advanced AI is delivered through centralized systems.

Tech entrepreneur and author Brett Hurt also called the U.S. order a “precedent.” In comments provided to Cointelegraph, he argued that the ability of a government to silence a commercial AI model overnight—without a public hearing, technical disclosure, or an appeals process—creates an “invisible ceiling” over labs operating in the country.

Both perspectives reinforce Grayscale’s central claim: when frontier AI is treated as a permissioned resource, policy decisions can instantly determine its availability. For developers and users, that raises practical questions about continuity, portability of workflows, and the feasibility of building systems that can operate across changing access regimes.

Looking ahead, investors and builders will likely watch whether decentralized AI networks see sustained usage (not only short-term token volatility) and whether policymakers clarify how cross-border access to frontier models will be handled in the future. The larger uncertainty is how long access constraints remain and whether similar directives spread to other model providers.

Bernstein has projected Robinhood’s prediction market revenue to reach $586 million in 2026, up from $150 million in 2025, as World Cup trading activity has pushed daily market volumes as high as $4.8 billion.

According to a Monday client note from research and brokerage firm Bernstein, prediction markets have become Robinhood’s fastest-growing revenue-generating product since launch, supported by a surge in trading tied to the FIFA World Cup.

The firm said daily prediction market volume climbed from $2.2 billion on June 11 to $4.8 billion on June 12, when the U.S. faced Paraguay. Bernstein noted that those figures already exceed the $1.4 billion traded during last season’s Super Bowl, one of the sector’s biggest events.

Based on its estimates, Robinhood’s prediction market business could contribute about 17% of transaction-based revenue and 10% of total company revenue in 2026.

World Cup activity drives prediction market growth

In its analysis, Bernstein pointed to Robinhood’s partnership with Rothera, a CFTC-licensed exchange and clearinghouse, as a major advantage. Since launching on May 28, Rothera has processed about 200 million contracts over its first 18 days, with FIFA World Cup and Major League Baseball contracts accounting for nearly all trading activity.

Bernstein said Robinhood’s distribution network gives it a competitive position in the sector. The firm cited the company’s large retail user base, a $0.01 commission per contract, and fee discounts of up to 50% for Gold subscribers.

Competition has also expanded as more firms introduce new contract categories. Bernstein highlighted Polymarket’s recent rollout of private company event contracts and Kalshi’s launch of CFTC-regulated perpetual futures tied to major cryptocurrencies. According to the firm, Kalshi’s crypto futures products generated $1 billion in trading volume within a week of launch.

Growing interest in private market exposure has coincided with activity elsewhere in the financial and crypto sectors. Earlier this month, Robinhood chief executive Vlad Tenev said Robinhood Securities received approval to act as an underwriter, allowing the company to participate directly in bringing firms public rather than only distributing IPO shares.

At the same time, crypto trading platforms have expanded products linked to private companies before they reach public markets. A report published by Talos and Coin Metrics last week said pre-IPO perpetual futures are increasingly being used as price discovery tools ahead of public listings. 

The report cited billions of dollars in trading volume tied to SpaceX-related contracts on Hyperliquid and said Cerebras Systems contracts traded within about 1% of the stock’s eventual opening price after listing.

Bernstein previously identified Robinhood, DraftKings and Coinbase as public companies positioned to benefit from rising prediction market participation during the World Cup. The firm estimates the tournament could generate more than $3 billion in additional handle and between $5 billion and $10 billion in extra consumer trading volume across the sector.

South Korea’s largest crypto exchange, Upbit, announced trading support for SPX6900 (SPX) on June 16. 

The exchange set SPX trading to start at 14:00 KST across KRW, BTC, and USDT markets. The listing gives SPX direct access to won-based trading and two major crypto pairs on one of the country’s main exchanges.

Bithumb also announced support for SPX6900 in the KRW market. The exchange set the SPX/KRW market to open at 17:00 KST, three hours after Upbit’s planned start. The two listings place SPX in front of Korean retail traders during the same trading day.

SPX6900 is a meme token based on a parody of the S&P 500. The project uses market culture and internet humor as its main identity. The listings do not change the token’s core use case, but they add new centralized exchange access in a major Asian crypto market.

Bithumb also adds Spacecoin

Bithumb also listed Spacecoin (SPACE) in the KRW market. The exchange set SPACE trading to start at 14:00 KST, with deposits and withdrawals expected within two hours of the notice. Bithumb said it supports SPACE deposits on the Ethereum network only.

Spacecoin is a DePIN project focused on satellite-based global internet access. Bithumb described the project as building a decentralized connection layer for global data transfer. The project’s token, SPACE, is expected to support satellite services, staking, and partner-related payments.

Advertisement

Spacecoin (SPACE) traded near $0.0077 at press time, indicating 17% increase in the past 24 hours and almost 20% in the past 7 days, according to CoinGecko data.

The SPX and SPACE listings show that Bithumb added two different token categories on the same day. SPX sits in the meme coin sector, while SPACE sits in the decentralized physical infrastructure network sector. Both opened with KRW market access, which lets Korean traders trade them directly against the won.

Bithumb also set trading controls for the new markets. The exchange said buy orders would be blocked for the first five minutes after trading starts. It also said some sell orders and order types would face limits during the early trading window.

The exchange also warned users about risk. Bithumb said virtual assets are “high-risk products,” and users can lose all or part of their funds. It added that investors remain responsible for their own trading decisions.

Advertisement

SPX price rises after listing news

Crypto.news price data showed SPX6900 trading at $0.377031 on June 16. The token rose 9.32% in 24 hours, while seven-day performance stood at 26.83%. Its 24-hour trading volume reached $27.69 million, with the price moving between $0.33316 and $0.39646 during the same period.

The token held a market capitalization of about $350.9 million and ranked #130 by market cap on the crypto.news price page. Its fully diluted valuation matched the same figure. The circulating supply stood at 930.99 million SPX, with a maximum supply of 1 billion tokens.

The latest daily chart showed a short-term recovery from a lower consolidation range. SPX has moved mostly sideways after a long decline from earlier highs. The current price remains well below its all-time high of $2.27, reached on July 28, 2025.

SPX was still down 74.8% over the past year, based on crypto.news data. The 30-day move stood at 1.59%, while the 200-day change was down 45.82%. These numbers show that the latest rebound came after a long pullback.

Advertisement

Momentum improves, but resistance remains

The short-term chart showed SPX moving back toward the upper part of its recent trading range. The nearest resistance area sits around $0.40 to $0.45, where price had faced selling pressure in May. A clean move above that zone would show stronger short-term demand, while another rejection would keep the range in place.

The relative strength index stood at 60.81, above its moving average of 44.78. That reading shows stronger buying pressure in the short term. The indicator remained below the overbought zone, so the move had not reached an extreme level by that measure.

The MACD also showed early improvement. The histogram was positive near 0.0082, while the MACD line stood above the signal line. Both lines remained close to the zero area, which shows that the rebound was still developing.

The Korean exchange listings added a new trading event for SPX6900 after a period of weak long-term performance. Traders will now watch whether Korean market access can support volume beyond the opening sessions. SPACE will also face its first KRW market test on Bithumb as users assess demand for the satellite-based DePIN project.

The listings also arrived during broader interest in Korean won trading pairs this week.

Advertisement

Disclosure: This article does not represent investment advice. The content and materials featured on this page are for educational purposes only.

Advertisement

XRP finally broke through the $1.20 level that had capped rallies for weeks, but buyers couldn’t keep control of the move.

After climbing as much as 10% and briefly trading near $1.25, the token ran into profit-taking that pushed it off session highs, putting the focus back on whether the breakout can hold rather than how far it can extend.

News Background

• XRP ETFs recorded a second straight week of inflows, attracting $10.68 million and lifting cumulative inflows to roughly $1.44 billion.

• South Korea’s Upbit exchange accounted for 31% of XRP wallet-flow activity by June 14, up from 13% a week earlier, highlighting strong regional demand.

• Ripple continued expanding its payments infrastructure through integrations including OpenPayd and RLUSD-related settlement activity.

Price Action Summary

• XRP climbed from roughly $1.14 to a session high near $1.25 before pulling back.

• The breakout was driven by a volume surge that reached more than 180 million XRP, easily clearing resistance around $1.20.

• Selling emerged near $1.25, trimming gains and leaving traders focused on whether former resistance can now hold as support.

Technical Analysis

• The move confirmed a breakout from the early-June consolidation range.

Nvidia is reportedly preparing a major debt raise aimed at funding artificial intelligence infrastructure, a move that further highlights how capital markets are backing the AI buildout—even as some crypto-native miners search for ways to use the same power and data-center assets for non-crypto computing demand. Bloomberg reported on Monday that Nvidia is seeking to raise at least $20 billion through a multi-part bond offering, which would finance AI-related investments and refinance existing debt.

The reported structure points to the scale of Nvidia’s ambitions and the continuing appetite from investors for long-dated, “high-grade” corporate financing tied to AI growth. For the crypto industry, the development matters because it reinforces the broader trend of miners pivoting toward AI hosting and high-performance computing—an area where electricity, cooling, and server capacity can be leveraged beyond Bitcoin.

Key takeaways

• Nvidia is reportedly targeting at least $20 billion in a multi-maturity bond sale to fund AI investment and refinance debt, according to Bloomberg.
• The longest-dated notes are expected to price around 0.9 percentage points above comparable U.S. Treasury yields, per the report’s description of expected pricing.
• Nvidia’s GPU dominance makes its capital spending plans a key barometer for AI infrastructure demand across hyperscalers and cloud providers.
• Bitcoin miners have increasingly looked to AI hosting and data-center services as Bitcoin mining economics face margin pressure after the April 2024 halving.
• Industry data cited in the article suggests miners have been selling significant portions of their BTC holdings, aligning with a push to diversify revenue streams.

Nvidia’s reported $20 billion debt plan reflects AI infrastructure demand

Bloomberg, citing people familiar with the matter, said Nvidia is kicking off its first high-grade bond offering since 2021 and plans to sell notes across seven maturities ranging from two to 30 years. The report also described pricing expectations for the longest-dated bonds, indicating yields roughly 0.9 percentage points above comparable U.S. Treasuries.

While the immediate story is about corporate financing, the underlying message is about what investors are willing to underwrite: companies at the center of AI supply chains that can convert demand for computing power into sustained revenue. Nvidia’s GPUs are widely used to train and run large language models, and its role ties capital expenditures by cloud providers and AI users directly to the semiconductor ecosystem.

That makes Nvidia’s bond plans more than routine fundraising. When a major infrastructure supplier raises substantial long-term capital for AI initiatives, it signals confidence that demand for accelerated computing, data-center buildouts, and related capacity will persist long enough to justify multi-year funding.

Miners are repurposing power and data-center assets for AI

Bitcoin mining has historically depended on energy-intensive operations and the ability to monetize block rewards and transaction fees. As competition and operating costs rise, some operators have pursued a different angle: using existing power infrastructure and data-center capabilities to support AI hosting and other high-performance computing workloads.

According to the article, companies that previously relied heavily on Bitcoin mining revenue—including HIVE Digital, TeraWulf, Hut 8, and CleanSpark—have been positioning themselves to provide data-center capacity. The core logic is straightforward: if a miner already has power agreements, cooling systems, and space designed for constant compute workloads, it can redeploy that infrastructure toward customers needing compute resources outside of crypto.

That shift also reflects a sector-level asymmetry. AI demand is not constrained to the same cycles as crypto markets, while electricity remains a key input in both ecosystems. For miners with surplus or underutilized hosting capacity, AI can become a way to reduce dependence on the volatility of Bitcoin’s price and mining difficulty.

Why Bitcoin mining economics are pushing diversification

The pivot toward AI has accelerated as the economics of core mining have come under strain. The article notes that margin pressure has worsened following the April 2024 halving, when Bitcoin’s block reward was cut—an event that reduced revenue per unit of work while mining difficulty and operating costs remained challenging.

Analysts cited in the piece characterized the environment as unusually harsh, with many miners responding by trimming exposure: selling parts of their Bitcoin treasuries, reducing leverage, and looking for new revenue streams. These actions are consistent with businesses trying to stabilize cash flow while the “payoff per share of hashrate” is less generous than it was pre-halving.

Supporting data referenced in the article comes from TheEnergyMag. It states that Bitcoin miners collectively sold more than 15,000 BTC between October and March. The report further describes an acceleration after October, when BTC peaked above $126,000, suggesting that the cash and liquidity pressures did not ease as markets moved.

For investors, the operational implication is that mining companies are being forced to choose between competing priorities: maintaining mining activity while retooling business models to monetize the same infrastructure through AI- and data-center-adjacent services.

AI infrastructure expectations are spreading beyond semiconductors

Nvidia’s bond plan underscores that traditional and AI-native financing channels are funding the AI compute buildout. But for crypto miners, the question is whether they can translate their infrastructure into stable AI-related revenue at scale.

The article points to analyst expectations from Bernstein, which reportedly expects IREN to derive the vast majority of its value from AI infrastructure rather than from Bitcoin mining. While that specific forecast is tied to a single company, it reflects a broader industry narrative: miners are increasingly evaluated not only as holders of mining assets, but as potential operators of compute capacity.

That shift also changes how market participants assess risk. Bitcoin mining revenue depends on a combination of Bitcoin price, network difficulty, and operational efficiency, while AI hosting revenue depends more on customer demand, contract terms, and the ability to deliver usable compute capacity competitively. Nvidia’s financing move, meanwhile, suggests the upstream supply chain for AI infrastructure is continuing to attract capital—an important backdrop for miners trying to secure demand.

As these stories converge, readers should watch for two practical indicators: whether miners secure meaningful AI hosting contracts (and at what margins), and whether debt and funding costs remain manageable as operators continue to refinance, reposition, or reduce exposure to crypto-linked balance sheet risk.

In the near term, Nvidia’s reported bond issuance will be closely monitored as a signal of AI capex momentum, while the next phase of the miner story will likely hinge on how quickly real hosting demand materializes and whether diversification can offset ongoing pressure in Bitcoin mining economics.

BitMine Immersion Technologies has continued adding to its Ethereum treasury holdings, purchasing a further 76,881 ETH over the past week despite a broader market slump. The incremental buys come as Ether recently dipped toward the $1,600 area, highlighting how the company is maintaining a steady accumulation strategy rather than waiting for a rebound.

In its latest update, BitMine reported that it now holds 5,620,754 ETH at an average acquisition price of $1,718. At the time of reporting, Ether was trading at $1,843.69, according to CoinMarketCap, placing the portfolio at roughly $10.2 billion in value. However, DropsTab data cited by the company’s update indicates the holdings still carry unrealized losses of nearly $9 billion relative to the average cost basis.

Key takeaways

• BitMine acquired 76,881 ETH in the past week, bringing total holdings to 5,620,754 ETH at an average price of $1,718.
• The treasury position is valued at roughly $10.2 billion at reported prices, but DropsTab data estimates unrealized losses near $9 billion.
• BitMine controls about 4.66% of Ether’s circulating supply, moving closer to its stated goal of owning 5% of the 120.68 million ETH in circulation.
• The company has staked more than 4.1 million ETH, generating ongoing protocol rewards that may help offset price volatility.
• Broader Ethereum headwinds include spot ETF outflows and questions around how layer-2 adoption affects mainnet fee burn and deflationary dynamics.

Steady accumulation in a weak tape

BitMine’s latest purchase extends a pattern of consistent Ether buying throughout the bear market. The company’s update notes that the week’s acquisition period may have included moments when ETH briefly traded below $1,600, according to Cointelegraph’s reference to market conditions during that time.

While the move has helped narrow BitMine’s average cost basis, the scale of its position means the overall portfolio remains exposed to large unrealized drawdowns. Even with Ether above the $1,700 average reported cost, DropsTab’s figures—referenced in the update—suggest losses are still substantial in absolute terms.

From an investor perspective, the key signal isn’t just the size of the buy, but the decision to continue accumulating during downturn conditions. Treasury-style strategies typically aim to reduce timing risk, yet they also require patience as mark-to-market losses can remain significant for extended periods.

Approaching the 5% supply target, with staking underneath

BitMine said its growing holdings bring it closer to a long-stated objective: owning 5% of Ether’s total circulating supply of 120.68 million tokens. Based on its current position, the company controls approximately 4.66% of all ETH.

In parallel with its spot accumulation, BitMine has staked more than 4.1 million ETH. Using the prices cited in the update, that staked amount is worth roughly $8.1 billion. Staking supports the Ethereum network by helping secure consensus and enables the company to receive protocol rewards, creating a recurring source of yield that can continue even when ETH prices weaken.

This matters because treasury models with staking components can partially decouple “yield generation” from “price appreciation.” Even if Ether’s market value declines, staking rewards may provide incremental performance, though they also come with staking-specific risks and lockups inherent to the system.

ETF outflows and their pressure on demand

Ethereum’s challenges this year are not limited to spot market weakness. The downturn has also weighed on spot Ether exchange-traded funds, which recorded four consecutive days of net outflows last week. CoinShares-style performance measures can vary by provider, but the article’s figures point to persistent selling pressure, including days where net outflows exceeded $60 million.

BlackRock’s iShares Ethereum Trust ETF (ETHA) remains the largest US-listed ETH ETF. According to the update referencing SoSoValue, ETHA has net assets of $4.75 billion and holds 2.36% of the crypto’s circulating supply. In practice, continued ETF outflows can reduce incremental, regulated demand at exactly the moment spot liquidity is most sensitive to broader risk appetite.

For market participants, the tension is clear: treasury buyers like BitMine may be absorbing supply, but ETF flows reflect how traditional investors are responding to uncertainty around Ethereum’s longer-term economics and growth trajectory.

Layer-2 adoption and Ethereum’s fee-burn debate

beyond ETF flows, the update highlights structural questions about Ethereum’s future revenue and deflation dynamics. Ethereum’s layer-2 scaling strategy is designed to move more transaction activity off the main chain, improving speed and lowering costs for users.

However, as more activity migrates to layer-2 networks, the Ethereum mainnet captures less transaction-fee revenue. That can also reduce the amount of ETH burned by the protocol, weakening the mechanism that has historically contributed to deflationary pressure.

The result is an ongoing debate: while layer-2s may support overall ecosystem usage and liquidity, they can alter the mainnet’s cash-flow and supply dynamics that investors track. If the majority of activity shifts away from the base layer without a corresponding economic balancing mechanism, long-term holders may need to underwrite Ethereum’s value proposition on factors beyond native fee burn.

Foundation leadership departures add governance uncertainty

In addition to market and protocol-level questions, the update points to internal governance and organizational changes. It states that at least nine senior leaders, researchers, and core contributors have departed the Ethereum Foundation this year, characterizing the wave as one of the largest talent attrition events in its history.

The departures are described as occurring alongside an organizational overhaul and renewed community debate over the foundation’s governance, strategic direction, and role in Ethereum’s long-term development. Even when such moves do not immediately change protocol code, they can influence investor sentiment by affecting expectations around coordination, research priorities, and how quickly contentious issues are resolved.

For readers watching the sector, this is a reminder that Ethereum’s narrative is shaped not only by technical scaling, but also by institutional capacity and how decisions are communicated and managed across the community.

Going forward, the market will likely track whether BitMine’s accumulation and staking yield can continue to offset the portfolio’s unrealized losses, while the broader ecosystem watches ETF flow trends, the pace of layer-2 migration, and any further clarity—or lack thereof—around Ethereum Foundation direction.

Strategy has added 1,587 BTC to its balance sheet, two weeks after its first disclosed Bitcoin sale in years raised questions. 

Michael Saylor said on X that the company bought the coins for about $100 million. Saylor had hinted at the move hours earlier with a short post saying, “Another Orange Star,” a phrase he often uses before Strategy Bitcoin updates.

Saylor wrote, “Strategy has acquired 1,587 BTC for $100 million,” adding that the purchase lifted the firm’s Bitcoin reserve to 846,842 BTC. Strategy also increased its U.S. dollar reserve by another $100 million to $1.1 billion.

Binance Research said April’s DeFi exploits triggered about $13 billion in total value locked outflows, cutting liquidity across on-chain protocols. 

The research arm said the on-chain leverage ratio rose to about 38%, a level last seen in 2021, as TVL fell faster than borrowing.

The move did not come from a clear return in real borrowing demand. Binance Research said “meaningful deleveraging has yet to materialize,” even after a wider crypto market pullback. That means the ratio moved higher because the base of locked capital became smaller. When TVL falls, each dollar of debt weighs more on the system.

Drift and KelpDAO drive April losses

Binance’s May market report said DeFi TVL fell 10.7% month over month to $82.7 billion in April. It also said protocols suffered $635.24 million in exploits during the month, the highest monthly total since the Bybit incident in February 2025. DefiLlama counted 28 hack events during April, which Binance called a record monthly count.

As crypto.news reported, the first 18 days of April already saw more than $606 million stolen across 12 incidents. The two largest attacks were Drift Protocol, at about $285 million, and KelpDAO, at about $292 million. Later, crypto.news reported that the two attacks together represented $577 million in losses and were linked to North Korea’s Lazarus Group.

Those two cases carried most of April’s reported losses. They also showed that DeFi exploit risk no longer comes only from code bugs. Reports tied the attacks to social engineering, compromised systems, governance weaknesses, and bridge infrastructure.

Aave and KelpDAO recovery stay in focus

The KelpDAO incident also spread pressure across connected lending markets. Binance Research said the KelpDAO exploit created about $230 million in bad debt on Aave and cut Aave’s TVL by half. The event showed how one bridge failure can move through DeFi when stolen collateral enters lending markets.

KelpDAO later completed the operational part of its rsETH recovery plan. As previously reported, the protocol sent a final batch of 20,373.7 rsETH to the LayerZero smart contract used for cross-chain transfers. The protocol said minting, redemptions, and reward functions were operating normally again after earlier restart steps.

The recovery steps reduced some direct pressure on KelpDAO users. They did not remove the wider concern around DeFi leverage. Binance Research’s data suggests that the market still carries debt against a smaller pool of locked assets.

Recent exploits show risks remain active

Security incidents continued after April, though reported losses dropped in May. CertiK put May hack losses at $68.3 million, down nearly 90% from April’s roughly $650 million, as reported. Still, DeFi projects kept facing attacks tied to bridges, old contracts, private keys, and operational controls.

Recent cases include Humanity Protocol, Aztec Connect, and Raydium. Humanity Protocol said more than $36 million was stolen after attackers compromised administrative keys linked to its bridge systems. Aztec Connect lost about $2.1 million from an old immutable contract, while Raydium said it would reimburse users after a $1.3 million exploit hit five legacy Solana liquidity pools.

The latest cases keep DeFi security in focus as leverage remains elevated and liquidity remains weaker than before April’s exploit wave. Binance Research’s reading points to a market where TVL has fallen, borrowing has not recovered strongly, and deleveraging remains incomplete.

Bitcoin BTC$65,790.78 briefly traded above $67,000 late Monday before slipping back under $66,000 in a move that is indicative of how cautiously crypto is treating the Iran peace deal that has rallied other markets.

The token changed hands at $65,845 on Tuesday, up 0.3% over 24 hours and 4.8% on the week, per CoinDesk data. It touched a 24-hour high of $67,217 before fading. Ether held up better, rising 2.8% on the day to $1,764 and 5.8% on the week. Solana gained 3.2% to $73, XRP added 3.2% to $1.22 and Hyperliquid’s HYPE led the majors again, up 6.3% to $69.

The macro backdrop turned sharply friendlier on Monday. President Donald Trump and Vice President JD Vance signed an electronic copy of a memorandum of understanding with Iran, and Trump said the Strait of Hormuz, already partially open, will fully reopen on Friday.

Brent crude slipped below $83 a barrel after its biggest drop in more than two weeks. The S&P 500 added 1.7% on Monday and the Nasdaq 100 rose 3.1%.

Yet bitcoin has not moved like an asset pricing in relief.

Grayscale researchers say Anthropic’s abrupt shutdown of access to its latest frontier AI models following a US government directive underscores the risks of centralized control over advanced AI systems. In a Monday note, Grayscale head of research Zach Pandl argued that the episode could accelerate interest in decentralized alternatives such as Bittensor.

According to the report, the US ordered Anthropic to suspend access to its models for foreign nationals on national security grounds. Anthropic then disabled access to Fable 5 and Mythos 5 for all users to comply with the directive, prompting a measurable shift in crypto market attention toward decentralized AI networks.

Key takeaways

• Grayscale’s Zach Pandl links Anthropic’s compliance move to the broader problem of centralized “frontier AI” access being controlled by a small number of entities.
• The US directive focused on foreign nationals, but Anthropic disabled access for all users, which Pandl called a warning sign for access risk.
• Grayscale reports that TAO rose sharply after the cut-off, climbing 30% within 12 hours and reaching a three-week high of $283 on Monday.
• Bittensor is positioned as an alternative network intended to provide AI access through decentralized infrastructure rather than a single lab.
• Industry observers cited by Cointelegraph argue the event sets a precedent for how governments can restrict commercial AI models quickly, potentially without standard procedural safeguards.

US directive prompts a wider shutdown

Cointelegraph reported that on Friday the US government directed Anthropic to suspend access to its AI models for foreign nationals, citing national security concerns. In response, Anthropic disabled access to Fable 5 and Mythos 5 for all users, not just those affected by the foreign-national requirement.

Pandl pointed to the speed and breadth of the change as evidence that centralized frontier AI access can be constrained overnight. He framed the episode as more than a policy dispute: it is a practical demonstration of how quickly access to cutting-edge capabilities can be revoked when decision power sits with a small set of institutions.

Grayscale: centralized control drives demand for decentralized AI

In his Monday note, Pandl said the US order “shows the centralized control of frontier AI technology and drives home the need for decentralized alternatives.” He argued that investors are likely to keep looking for different architectures that don’t rely on one company’s ability to grant or suspend access.

Grayscale expects that demand for decentralized AI—citing Bittensor specifically—will continue to rise as users search for options that are not subject to the same access chokepoints. Pandl linked this to the idea that governments and large AI labs increasingly influence “who can access these tools and under what conditions,” particularly as AI capabilities advance.

To illustrate the market reaction, Grayscale said that in the 12 hours after Anthropic cut access to its latest models, Bittensor’s TAO token climbed 30%. The note also claims TAO reached $283, a three-week high, on Monday—an indicator that traders were actively repricing decentralization narratives in response to the event. (TAO performance and the cited price level were attributed in the source to CoinGecko.)

“Think of it as Bitcoin for AI.”

Pandl described Bittensor as aiming to provide access to AI resources through an open, global, decentralized network—an “alternative vision” meant to reduce reliance on a single provider or centralized permissioning.

Why investors are watching decentralized networks

The debate here is not only technical; it’s about resilience. When a model vendor disables a service, users can lose access regardless of their location, and builders may have less certainty about continuity. Grayscale’s framing suggests that centralized AI deployment increases the probability of sudden disruptions tied to regulatory or security directives.

For market participants, the takeaway is that decentralized AI ecosystems are being evaluated not just on model quality or tooling, but on the structure of access itself. In other words, the episode became a live stress test of how quickly frontier AI access can change—and that test appears to have influenced attention toward networks positioned as alternatives.

However, important uncertainty remains: decentralized networks do not automatically guarantee immunity from regulation or other forms of restriction, and crypto token performance can reflect multiple factors besides the specific access event. Still, the timing described in Grayscale’s note suggests that traders and holders interpret the Anthropic directive as supportive of decentralization narratives.

Industry voices call it a precedent for AI governance

Beyond Grayscale, the source also includes comments from other participants in the AI-and-crypto space. Cointelegraph quoted EdgeRunner AI co-founder Colton Malkerson, who argued that the incident marks a “breaking point” for corporate data independence. He compared centralized AI access to “renting” intelligence from big labs, saying it is worse when access can be canceled and the provider can monitor the user’s activities as a condition of the service.

Tech entrepreneur and author Brett Hurt likewise described the US action as “a precedent,” arguing that if a government can silence a commercial AI model overnight without public hearing, technical disclosure, or an appeals process, then all labs may effectively operate under an unseen constraint.

These viewpoints align with Grayscale’s central message: access to advanced AI is increasingly treated as a policy lever. For crypto-native AI networks, that creates a motivating question for investors and users—whether decentralized systems can offer more continuity when centralized providers face sudden external directives.

Going forward, readers should watch how Anthropic’s compliance approach evolves—particularly whether access remains uniformly disabled—and whether additional policy moves target other frontier model providers. At the same time, market participants will likely continue tracking whether decentralized AI tokens capture sustained inflows, or whether the initial reaction fades as the situation clarifies.