
This week, “Deconstruction” focuses on U.S. crypto regulation, the AI race slowdown, Bitcoin’s reaction to the Fed rate, a Revolut data breach, and external access to ChatGPT conversations.
U.S. Crypto Regulation
On September 15, the CLARITY Act failed a procedural vote in the Senate: 49 senators voted in favor, while 50 opposed. The market may have already priced in the failure: from July 1 to September 15, the likelihood of the bill’s passage on Polymarket dropped from 39% to 18%, while Bitcoin rose from $58,000 to $80,000 during the same period.
The next day, a House committee approved a bill to establish a strategic Bitcoin reserve. Selling government-held Bitcoin is prohibited for 20 years. This could have more long-term significance for the market than the Senate vote failure.
AI Race Hits Safety Concerns
Following calls from Dario Amodei, Sam Altman, and Elon Musk to cautiously develop advanced AI models, shares of major chipmakers fell. For investors, this signals a potential slowdown in spending on computing power, memory, electricity, and data centers. Meanwhile, Jensen Huang opposed new AI regulatory requirements, while Anthropic and OpenAI supported independent audits.
OpenAI also disclosed six instances of inappropriate model behavior: systems concealed errors, used API keys without permission, and uploaded files to external services against restrictions. The company is implementing continuous monitoring. As AI agents gain more autonomy, the need for oversight of their actions increases.
Bitcoin Withstands Rate Hike
On September 16, the Fed raised the key rate by 25 basis points to 3.75–4%. Bitcoin rose by 1% and remained above $76,000, as the market had anticipated the decision.
Analysts have mixed views on future trends: some see signs of the end of a bear phase, while others predict a shift into a bear zone. Bitcoin ETFs saw $745 million in outflows over two days, but on September 17, funds attracted about $159 million. The weekly flow remains negative.
Revolut Data Breach
Revolut confirmed a data breach affecting approximately 680 customers after employees provided confidential information in response to fake requests. The company’s systems were not hacked, and customer funds were not affected, but the perpetrators obtained personal and banking data, as well as transaction histories.
On September 16, the group iamnotavillain demanded 6000 XMR—about $3 million—and threatened to sell the data to other criminals. Among those affected was former Mt. Gox head Mark Karpeles. The main threat now involves potential fraud and new attacks using the stolen information.
External Access to ChatGPT Conversations
Hundreds of OpenAI contractors gained access to real user queries and ChatGPT conversations to evaluate model responses. These data were used to make the model less likely to adapt to the user and agree without necessity.
Some conversations contained personal and sensitive information, and the Privacy Filter could miss certain personal data. New ChatGPT conversations for Free, Plus, and Pro users are used for model training by default, but this setting can be disabled. It’s important to consider these parameters before sending sensitive information.
This is a shortened version of the podcast. Watch the full episode:
Subscribe to the podcast:
Apple Podcasts
Spotify
YouTube
Deezer
Yandex Music
YouTube Music
