Anthropic secretly lobbied Congress, advocating that the government can only shut down an AI model if it poses a risk of more than $1 billion in losses or at least 50 casualties.

Beating AI News Flash: DeepSeek apparently rolled out an update to its model version V4.1-Flash-1017 today. Per community user feedback, the model identifier deepseek-v4.1-flash-1017 is now accessible on the API platform, though the official has not yet announced specific update content or performance metrics for this version, with further details still pending confirmation.

Beating AI News Flash via Bloomberg: Jump Trading and DRW are ramping up their positions in the AI compute power-linked derivatives market, planning to trade compute futures and other contracts to help enterprises hedge against GPU rental price volatility. Both firms entered the cryptocurrency market early, and now aim to replicate their market-making experience there. Jump has set up a dedicated trading team, will use its own balance sheet to take on trading risks, and plans to provide market-making services for compute futures being prepared by institutions like CME and Silicon Data, while also monitoring related products from ICE and Ornn. DRW, meanwhile, plans to participate in compute trading via three teams, and has already invested in Silicon Data and Ornn. CME’s relevant futures products still await regulatory approval. This push comes amid growing skepticism toward AI infrastructure investments. Stocks and bonds of compute firms like CoreWeave are under pressure, as markets reassess whether AI revenues can support data center construction costs. Boston Consulting Group projects the global AI compute market could reach $2.3 trillion by 2030. However, the compute derivatives market remains in its early stages, with insufficient price transparency. GPU rental prices vary widely by region, configuration and supplier, which may cause futures indices to diverge from enterprises’ actual procurement costs. Additionally, compute power cannot be stored like oil or copper, and unused GPU compute time cannot be reserved for future use. Whether Jump and DRW can turn compute power into a tradable, hedgeable standardized asset remains to be seen.

Former US President Trump posted a graphic on his social media platform, claiming that during his tenure, the S&P 500 surged from 5,881 points in December 2024 to 7,651 points in September 2026, a cumulative gain of roughly 30%. He captioned the post: "The strongest in history!" However, recent risks tied to AI trading have pressured US equities, which fell broadly on Friday. Concerns over AI investment returns, paired with soaring international oil prices, triggered fresh sell-offs in tech stocks. The Nasdaq Composite closed down 2.15%, the Nasdaq 100 dropped 2.5%, and the S&P 500 slipped 1.1%. The Nasdaq notched a 4.48% weekly loss, while the Dow Jones Industrial Average rose 0.68% — their weekly performance gap was the largest since July 2002. The Philadelphia Semiconductor Index declined for the eighth straight trading day, marking its longest losing streak since 2008. Earlier today, Trump delivered his latest comments on the Russia-Ukraine and Middle East issues: "I know some details of the Russia-Ukraine conflict, and I will make them public at some point. I hope it ends. The Biden administration did a lot of bad things regarding the Ukraine issue." When asked about inflation, Trump again asserted: "Oil prices will drop soon. Israel's support for Saudi Arabia is very substantial, and the Abraham Accords are back on Saudi Arabia's agenda."

According to Arkham's monitoring, whale trader 0x9c6 holds a 20x leveraged long position on SOL on Hyperliquid, with the current position value totaling around $57.5 million. The entry price for the position is $80.86. Based on SOL's current price of $104.43, the position has an unrealized profit of roughly $13 million. The trader opened the position approximately one month ago and has not closed it to date. Their Hyperliquid account is currently valued at about $24.2 million.

Crypto market analyst CW noted in a post that Binance’s spot trading volume typically fluctuated between $6 billion and $13 billion before mid-August; it later rose to $19.4 billion at one point, and has recently held steady at around $10 billion daily. Following the previous rally in the crypto market, Binance’s overall spot trading volume has remained at a high level. Despite the current short-term bearish trend in the market, Binance’s trading volume remains strong, indicating that market liquidity is still relatively sufficient. In contrast, Coinbase’s trading volume rose significantly during the recent market upswing, but as the market entered a consolidation and correction phase, trading activity has gradually cooled down.

TRON founder Justin Sun recently shared his personal asset allocation views (excluding TRON) during a 3.5-hour face-to-face talk with Jin Ming. Key points of his core portfolio strategy: - BTC: Held as a base position, with no reductions over the past year. While he expressed concerns about potential impacts from quantum computing and AI algorithms, noting the market has not fully priced these risks, he still maintains his BTC holdings and uses cold wallets in "bunker mode" for isolation. - ETH: Last year, he cut roughly 80-90% of his aggressive ETH positions in the $3,600–$4,000 range, and now only retains about 20% as a base position for observation, with no repurchase plans. Most funds from the ETH position reduction were allocated to US stocks. - NVIDIA: Serves as his main US stock holding, positioned as a base in the AI sector, with a heavy stake that he has not reduced. - Nasdaq 100: Held with a heavy position. He holds a slightly bearish mid-term outlook (citing factors like capital clustering and high interest rates) but has not opened short positions, keeping the position intact. - A-share market: No holdings. Overall, Sun is bearish on traditional assets and more focused on areas like AI applications and Web3. During the interview, he emphasized ETH as an aggressive asset while BTC acts as a base position. He added that he is bullish on AI health management and other applications, and bearish on some hardware sectors. The above are key takeaways from the interview, not a complete balance sheet.