Fed minutes say AI buildout continues to outpace expectations
Minutes from the Federal Reserve’s September FOMC meeting showed that most participants saw another rate hike before year-end as potentially appropriate. Several participants said the current policy rate was either not restrictive or only mildly restrictive, while some said upside inflation risks had increased in recent months.
The minutes also said staff raised their outlook for the economy and employment. GDP growth was projected to accelerate in the second half of the year and remain above potential through 2028, while the unemployment rate was expected to stay below its longer-run level through 2029. Inflation forecasts for 2026 through 2028 were revised up from July, with inflation seen returning to 2% by 2029.
On artificial intelligence, several participants said the scale and pace of AI buildout continued to exceed expectations. Participants generally agreed that AI investment could lift productivity and potential output over the next few years, though there was substantial uncertainty over how large that effect would be and when it would show up.
U.S. weekly jobless claims come in below forecasts
Initial jobless claims in the United States for the week ending Oct. 3 totaled 197,000, below the 200,000 expected. The prior reading was revised to 199,000 from 197,000.
Vitalik Buterin warns AI could pressure post-quantum cryptography
Ethereum co-founder Vitalik Buterin said in a post that AI-driven acceleration in mathematical research could materially weaken the practical security of lattice-based cryptographic systems within the next two years. He said that risk extends to technologies including ML-DSA and fully homomorphic encryption, or FHE, and could even bring forward the timeline for threats to ECDSA.
Buterin said AI could deliver the equivalent of 50 years of mathematical research progress within two years. He added that Ethereum’s Lean roadmap has gradually shifted over the past year toward pure hash-based signature schemes to reduce reliance on lattice cryptography.
On asset protection, Buterin said users should, where operationally practical, prioritize storing funds in addresses that have not yet initiated transactions. At the same time, he said users should not hurriedly move assets right away because transfers themselves carry a risk of fund loss. He also said multisig wallets should use off-chain signature confirmation where possible so signature data is not publicly exposed.
Greece consults on a bill with a 10% tax on crypto capital gains
According to an announcement from Greece’s finance ministry, the government has started public consultation on a new crypto-asset tax bill. The proposal would levy a 10% tax on capital gains from individuals selling crypto assets, while annual gains of up to 500 euros would be exempt. Crypto-to-crypto swaps would not trigger capital gains tax under the proposal.
Income generated from staking, lending, and providing liquidity would be taxed at 10% as interest income. Investors would also be allowed to voluntarily declare prior crypto trading gains within 12 months after the law is published, and those meeting the conditions could have fines and interest waived.
CrowdStrike says AI-assisted attacks hit South Korean financial institutions
Security firm CrowdStrike said in an intelligence report that an unidentified hacker carried out targeted cyberattacks against multiple South Korean financial institutions from late September to early October, combining large language models, or LLMs, with the domestic open-source AI penetration testing tool ARTEX.
Based on analysis of an exposed open directory, the attacker used a dual-server setup centered on a Hong Kong IP address and coordinated the attack flow with Claude Code, GLM-5.3, Grok 4.6, and DeepSeek v4.1-flash accessed through an API proxy. CrowdStrike also said the attacker had asked Claude how to monetize leaked South Korean data in Telegram groups.
South Korean media estimated that at least seven financial institutions were affected, including KB Kookmin Bank, Shinhan Bank, and Hana Bank. Personal data tied to about 68,000 people was exposed, including annual income and loan-limit information.