Bitcoin Companies Want Help From AI Labs to Guard Against Hackers
More than 40 bitcoin and crypto firms requested the largest AI labs to allow independent security researchers access to their strongest models before public release. The letter, published on August 10, highlights that the Bitcoin network secures assets worth over a trillion dollars, with wallets and custody systems holding even more. The Bitcoin Policy Institute organized the open letter, which was signed by Coinbase, Block, BitGo, Blockstream, ARK Invest, and over 40 other firms. The letter emphasizes a practical gap where developers patching wallets and node software cannot access the best AI models, while attackers already do. Anthropic demonstrated that the same AI models facilitate "vibe hacking," enabling criminals to conduct live Bitcoin ransom attacks more cheaply. The letter requests AI labs to accept the risk of granting trusted access to qualified defenders despite concerns about potential leaks. The Bitcoin Security Consortium, including BlackRock, Coinbase, Anchorage, ARK, Blockstream, Fidelity Digital Assets, and Galaxy, pledged $15 million for long-term network security. However, the letter targets a different issue: providing current AI models to maintainers who patch vulnerabilities rather than funding future threats. The letter urges frontier AI labs to establish or expand trusted-access programs for qualified defenders of open-source financial infrastructure. It notes that labs and select partners have months-long early access to new attack capabilities, which eventually leak through public releases or compromised systems. Maintainers writing open-source code cannot join vetted researcher programs, and public models often restrict legitimate flaw-hunting due to guardrails. Currently, major AI labs have not publicly addressed this issue, but companies are working on safeguards to prevent malicious use of their models.
Bullish retains 19,990 Bitcoin worth $1.28B through Q2, cementing its treasury strategy
Bullish holds 19,990 Bitcoin valued at approximately $1.28 billion through the second quarter of 2026, maintaining a steady core position. The company's broader crypto treasury totals around $1.52 billion, with Bitcoin comprising nearly 98% of this total asset value. Bullish's Bitcoin holdings represent roughly 0.11% of the total Bitcoin supply, ranking it among the largest publicly listed corporate holders worldwide. Unlike typical corporate Bitcoin acquisitions, Bullish's Bitcoin was not bought through operating cash flow or equity raises aimed at crypto purchases. The Bitcoin position existed before Bullish's public listing and was incorporated into the company from its inception. Block.one, Bullish's parent, initially funded the exchange with an injection that included 164,000 BTC alongside other assets at its 2021 launch. This initial capital originated from Block.one's EOS token sale between 2017 and 2018, which raised approximately $4.1 billion. Therefore, Bullish's Bitcoin reserve is essentially a downstream product of a historic token sale predating the exchange's existence by several years. The 19,990 BTC figure reflects Bullish's retained core holding, while treasury trackers estimate the total Bitcoin position closer to 23,300 BTC. This discrepancy suggests that Bullish categorizes or reports its Bitcoin holdings across multiple accounts or layers. Bullish is the only crypto-native exchange listed on the NYSE, providing institutional investors with equity exposure to crypto market infrastructure. The Bitcoin treasury adds an investment dimension, making BLSH shares partially a bet on Bitcoin's long-term price trajectory.
Strategy, Metaplanet unrealized bitcoin losses highlight risk of concentrating on just one token
Tokyo-listed Metaplanet reported a paper loss of $1.5 billion on its 43,000 BTC holdings as of the end of June, highlighting significant unrealized losses. Strategy, the largest public digital asset treasury company, disclosed a comparable paper loss amounting to $8.2 billion on its bitcoin assets last month. Together, these losses approach nearly $10 billion, which if tokenized, would rank as the 11th largest digital asset by market value. This combined loss token would be valued just behind dogecoin and ahead of tokenized Treasury coins like ONDO, privacy coins such as ZEC, and DeFi giant AAVE. The concentration of risk in a single token is underscored by many digital asset treasury firms issuing debt to fund bitcoin purchases, raising concerns about financialization. Bitcoin lacks inherent yield, return, or cash flow, yet the market has not shown significant concern as BTC trades between $62,000 and $66,000 for weeks. Bitcoin's price has remained choppy in the low $60,000s since October last year, while the S&P 500 equity index surged to record highs during the same period. Some analysts remain optimistic that the bear market may have ended, noting Bitcoin's decline generally halted near $20,000, close to the 2017 bull market peak. Alex Kuptsikevich, chief analyst at FxPro, stated that the 2021 bull market peaks were close to current levels, supporting the view that bearish momentum is fading. The headline emphasizes that Strategy and Metaplanet's unrealized bitcoin losses highlight the risks of concentrating investments on just one token.
SharpLink Plans $200 Million ETH Allocation to Lido’s wstETH
SharpLink plans to allocate $200 million worth of ETH to Lido's wstETH token, representing about 106,000 ETH based on Kraken's ETH price of $1,889.84. As of August, SharpLink reported holdings of approximately 888,938 ETH, including native ETH, LsETH, and weETH tokens. SharpLink's ETH holdings as of June 28 included 632,719 native ETH, 181,299 ETH represented by LsETH, and 72,707 ETH represented by weETH. Native ETH is directly staked in Ethereum, while LsETH represents staked ETH plus network rewards and can be used in decentralized finance applications. The weETH token includes base staking rewards and restaking economics facilitated through EigenLayer, providing additional staking exposure. Lido's wstETH is a wrapped form of stETH with a fixed balance, reflecting accrued rewards and designed specifically for DeFi integrations. Lido's total value locked was about $17.9 billion, accounting for 50.6% of tracked liquid-staking TVL, with a supply APY of 2.2% at the time of data reading. The 2.2% supply APY measures Lido staking returns but does not include any additional returns or risks from deploying wstETH in other onchain strategies. SharpLink's second-quarter revenue was primarily from staking, totaling $11.2 million, while its liquid staking and restaking tokens caused a $76.1 million impairment. SharpLink's CEO Joseph Chalom stated that Lido's composability allows layering additional yield sources on top of ETH exposure and staking returns. The company will hold wstETH with Anchorage Digital, utilizing Lido's open-source liquid-staking middleware to add a widely used DeFi asset to its portfolio.
Chinese AI Beats Restricted OpenAI and Anthropic Cybersecurity Models, Bitcoin Industry Warns
Bitcoin company leaders and open-source developers report that Chinese AI models outperform restricted American frontier systems in defensive cybersecurity work. Rob Hamilton, CEO of AnchorWatch, described crippling American AI restrictions that force researchers to rely on Chinese AI to secure Bitcoin infrastructure. American models from OpenAI and Anthropic frequently refuse or restrict legitimate defensive work, even for users with explicit access, according to Bitcoin security researchers. Chinese models such as Kimi K3 operate without the same guardrails and deliver confirmed results in cybersecurity tasks for Bitcoin projects. PortlandHODL, a Bitcoin Core contributor, highlighted the performance gap between US-based frontier AI and Chinese open models in finding critical vulnerabilities. He expressed concern about relying on Chinese AI for security and called for OpenAI and Anthropic to create proper access programs for US defenders. Alex Thorn signed a Bitcoin Policy Institute open letter demanding trusted access to frontier AI models for open-source defenders in the US. The letter urges frontier AI labs to establish clear trusted-access programs, arguing current restrictions leave legitimate researchers without access to strong models. Signatories request early access to cyber-capable models, sufficient compute, secure code review environments, and direct channels with lab security teams. Francis Pouliot described how a Chinese open-source model identified and helped patch a money-stealing exploit, while American models refused to review the patch. Hamilton stated he will return to using Chinese open-source models for Bitcoin infrastructure research despite patriotic reservations due to American AI restrictions. Concerns about Chinese model hosting as an attack vector can be mitigated since these models are open source and can run on American-hosted data centers.
Reader comments