BTC $69,015.08 -6.21%
ETH $1,932.99 -5.59%
BNB $634.71 -6.20%
XRP $1.00 -3.64%
SOL $78.29 -5.62%
TRX $0.3108 -1.95%
DOGE $0.0636 -5.10%
ADA $0.1683 -2.41%
BCH $216.63 -5.58%
LINK $9.73 -3.50%
HYPE $70.20 -2.67%
AAVE $110.62 -6.48%
SUI $0.7725 -5.35%
XLM $0.1429 -2.47%
ZEC $907.08 -7.16%
AAPL $356.68 +0.17%
AMZN $268.20 +0.72%
GOOGL $300.23 +1.58%
MSFT $521.07 -0.05%
META $625.40 -0.50%
NVDA $164.13 +1.85%
TSLA $398.55 +2.96%
SNDK $922.64 -3.21%
INTC $60.97 +0.64%
SPCX $140.76 +0.88%
MU $794.39 +0.73%
AMD $575.74 +2.43%
BTC $69,015.08 -6.21%
ETH $1,932.99 -5.59%
BNB $634.71 -6.20%
XRP $1.00 -3.64%
SOL $78.29 -5.62%
TRX $0.3108 -1.95%
DOGE $0.0636 -5.10%
ADA $0.1683 -2.41%
BCH $216.63 -5.58%
LINK $9.73 -3.50%
HYPE $70.20 -2.67%
AAVE $110.62 -6.48%
SUI $0.7725 -5.35%
XLM $0.1429 -2.47%
ZEC $907.08 -7.16%
AAPL $356.68 +0.17%
AMZN $268.20 +0.72%
GOOGL $300.23 +1.58%
MSFT $521.07 -0.05%
META $625.40 -0.50%
NVDA $164.13 +1.85%
TSLA $398.55 +2.96%
SNDK $922.64 -3.21%
INTC $60.97 +0.64%
SPCX $140.76 +0.88%
MU $794.39 +0.73%
AMD $575.74 +2.43%

event

All
Article
Flash

Anthropic applies zero-knowledge proof technology to AI reasoning to prevent model swapping

Anthropic has open-sourced FlashProve, bringing the commonly used zkSNARK (zero-knowledge proof) technology from the cryptocurrency field into AI model inference.This technology was originally designed to prove the validity of blockchain transactions without disclosing transaction details. Anthropic aims to use it to solve another problem: when users pay a high price for an AI model's API, how can they know if the service provider has secretly switched to a cheaper model?FlashProve allows service providers to first generate encrypted fingerprints for the model weights and then provide mathematical proof for the inference results. Users do not need to obtain the model weights or rerun the model; they can check whether the specified output comes from the previously committed set of weights. Claude was also involved in the development and optimization of this system.Anthropic has already completed experiments on Meta's 400 billion parameter model Llama 4 Maverick. For every 1024 tokens of context processed, the generated proof is about 19 MB, and verification on a regular CPU takes about 17 seconds.However, currently, each proof can only verify one output token and cannot validate an entire response. The GPU computation time required to generate the proof is also about 70 to 95 times that of regular inference. Whether it can be used to identify API service providers that secretly switch models in the future will depend on when the cost of proof generation can be reduced.

Zhao Changpeng: Binance quickly responded to the attack incident to prevent panic from spreading, and user trust in the platform is quite solid

Binance founder Zhao Changpeng stated in an interview, "After the attack incident, Binance quickly stepped up to clarify that user funds were not affected and user assets are safe. Although this is just a very simple statement, it effectively prevented panic from spreading further and avoided a chain reaction.Everyone understands that this incident will not have further spillover effects. Unlike previous cases where some platforms or lending protocols went bankrupt, ultimately causing users to bear losses, this time users will not be forced to bear related losses. This is fundamentally different from some past events, which often triggered second and third rounds of chain reactions. I believe that, first, this statement calmed everyone down. Secondly, Binance's ability to make such a commitment based on its own credibility also made users willing to believe this statement. After all, many fund flows and transaction records on the blockchain are public and transparent.However, this situation is different; users understand that Binance has sufficient financial strength to bear related losses. Binance has the SAFU (Secure Asset Fund for Users), a futures insurance fund, as well as very strong capital strength and a solid balance sheet. It is this ability to respond quickly, take responsibility proactively, and ensure that all users are protected and all losses are compensated that has gradually restored calm to the market.One interesting point: on the day of the incident, the decline in the cryptocurrency market was actually smaller than that of the tech stock market, a situation that is almost unprecedented. Typically, when the stock market declines, the crypto market often experiences a larger drop, as investors looking to hold cash tend to sell off crypto assets. But this time, even with such a severe hacking attack on the crypto industry, the market decline was smaller. I believe this indicates that investors overall remained quite calm, and it reflects the solid trust in the platform that has been gradually built over the years. I feel very proud of this."
app_icon
ChainCatcher Building the Web3 world with innovations.