Author: K33 Research
Compiled by: Shenchao TechFlow
Shenchao Introduction: The U.S. government transferred approximately 17,700 BTC to Coinbase Prime this week (including batches related to the Bitfinex case). K33 judges that most of it looks more like custody scheduling than selling; during the same period, OpenAI's mathematical results have reignited concerns about crypto signature security. For those watching BTC supply shocks and the "AI -> cryptography" narrative, the key things to watch are the pace of Bitfinex claim distributions and whether elliptic curves have been confirmed as weakened - currently there is insufficient evidence.
The U.S. government transferred approximately 17,700 to 21,500 BTC
The U.S. government transferred approximately 17,700 BTC to Coinbase Prime this week, including one transaction of 12,267 BTC related to the Bitfinex hack case. This reduced labeled holdings from 324,527 BTC to 306,795 BTC. If wallets that appear to be government-held but have not yet been labeled are included, about 21,500 BTC entered Coinbase Prime. These coins went through test transfers, one-time addresses, and were then consolidated into deposit addresses.
Most of them are probably not being sold. Since 2024, Coinbase Prime has been the custodian for the U.S. Marshals, so transfers there are a custody step. The Bitfinex-related coins are still stuck in the claims process, and the Department of Justice wants to return them in BTC. The "no-sale order" for the Strategic Bitcoin Reserve only covers coins that have been fully forfeited, while allowing returns to victims. This week, about 1,200 coins from FTX and HashFlare were also moving; these may be sold because victims may be compensated in cash or through remission procedures. They used the same deposit address as the Bitfinex batch, pointing to a Marshals-owned account.
About 15k BTC - out of about 109.7k BTC in Bitfinex claims - moved within two days, possibly suggesting preparations for distribution, or a ruling we have not yet seen. Bitfinex has committed to using at least 80% of net recovered funds to buy back LEO; once it receives the coins, this could mean selling up to about 85k BTC over 18 months, roughly 150 BTC per day.
Treasury withdraws crypto mixing rule
According to a notice published on Monday, FinCEN is withdrawing two old proposed crypto rules. One is a 2023 plan that would have labeled international crypto mixing as a "primary money laundering concern," which would have forced banks to report mixing transactions down to the wallet address and IP address level.
The other is a 2020 proposal that would have required identity verification and recordkeeping for transactions involving self-hosted wallets. FinCEN said the definition of mixing was too broad, could suppress legitimate privacy use, and would also impose heavy reporting burdens on institutions.
Robinhood adds $25 million in BTC to its balance sheet
Robinhood added $25 million worth of Bitcoin to its balance sheet, and Senior Vice President Johann Kerbrat said the move was intended to show the company's commitment to crypto. He acknowledged that the amount is small for a company with a market value of about $100 billion and will not significantly change Robinhood's direction. The move comes as other public companies continue buying: Strategy added 334 BTC last week and now holds 848,000 BTC.
The purchase fits a larger crypto push. Robinhood plans to launch perpetual contract trading for eligible U.S. users in the coming months; its Layer 2 chain has reached $1.05 billion in total value locked since launching in July. Kerbrat also said demand for its stock tokens has hit the volume cap under the SEC innovation exemption.
One-line quick view
Standard Chartered will expand crypto custody services in Singapore
Moody's assigns Sky Protocol a B3 rating as institutional interest in USDS rises
Coinbase Pro will return; after integrating Deribit, it will form Coinbase Global Exchange
Trending topics
AI math breakthrough reignites crypto security concerns
On October 6, OpenAI announced 377 mathematical discoveries produced by an advanced internal AI model, covering fields including algebra, number theory, and computer science. Many proofs were verified using Lean, a software tool for checking the logic of mathematical arguments. These discoveries demonstrated AI's improved ability to tackle advanced mathematics, but none involved breaking modern cryptography.
The release still triggered concerns in the crypto industry. Bitcoin and Ethereum rely on mathematical problems considered extremely difficult to solve. Their security is partly based on attackers being unable to derive a private key from a public key. Previously, discussion of these risks focused mostly on future quantum computers. OpenAI's progress raises another possibility: AI discovers shortcuts that allow ordinary computers to break protections previously considered secure.
Ethereum researcher Justin Drake warned on October 7 that AI-driven discoveries could potentially weaken the signature systems used by Bitcoin and Ethereum. In the worst case, he suggested a breakthrough could arrive in months rather than years. Drake urged the industry to prepare for what he called "bunker mode," recommending gradually moving funds to unused addresses whose public keys have not yet been exposed. Such addresses hide the public key until funds are spent, providing extra protection.
Ethereum co-founder Vitalik Buterin acknowledged the potential danger but did not recommend immediately moving funds. A poorly executed transfer could cause immediate losses, while the AI threat remains hypothetical. Buterin also warned that AI could discover weaknesses in newer cryptographic methods, including those developed to resist quantum computing. He believes the industry should give more consideration to hash-based security systems, which may offer stronger protection against unexpected mathematical discoveries.
However, several cryptography experts strongly questioned these warnings. Coinbase cryptography head Yehuda Lindell argued that there is no evidence AI has weakened elliptic curve cryptography - the mathematical foundation behind many digital signatures. He emphasized that solving difficult mathematical problems does not necessarily mean existing security assumptions are fragile. Lindell also pointed out that a real breakthrough would extend far beyond cryptocurrency and could threaten banking websites, software security, and other digital infrastructure. The disagreement reflects a broader debate: whether precautionary measures are justified when there is no actual evidence of a vulnerability.
These concerns also appeared during a broader period of market weakness. In the days after OpenAI's announcement, BTC fell from about $86,000 to $80,000, but this move should not be viewed in isolation. The Nasdaq also pulled back after challenging record highs, and risk assets weakened broadly, indicating that the broader market environment played an important role.
Drake has played a central role in highlighting long-term security challenges - first quantum computing, now AI. Raising awareness of these risks can help concentrate attention and momentum to develop stronger future security. However, there is currently no evidence that elliptic curve cryptography has been weakened, and OpenAI's mathematical breakthroughs do not constitute any immediate threat. For now, this debate is mainly about preparing for potential future challenges, not responding to a proven security risk.