Malicious Code Steals Cryptocurrency Wallet Credentials, State Department to Delete Old X Posts
WASHINGTON, D.C. - In a week marked by both technological advancements and security breaches, the dYdX cryptocurrency exchange suffered a significant setback as malicious packages on the npm and PyPI repositories compromised user wallets, while the State Department announced it would delete all pre-January 20, 2025, posts from its X accounts. These events, alongside fluctuating Bitcoin prices and breakthroughs in artificial intelligence, shaped the news cycle.
Researchers from security firm Socket revealed that open-source packages published on npm and PyPI were laced with code designed to steal wallet credentials from dYdX developers and backend systems. The compromised packages, including versions of dydxprotocolv4-client-js, put all applications using them at risk, leading to complete wallet compromise and irreversible cryptocurrency theft, according to Ars Technica. The attack targeted both developers testing with real credentials and production end-users.
Meanwhile, the State Department confirmed to NPR that it would remove all posts from its public accounts on X made before President Trump returned to office on January 20, 2025. Older posts will be archived internally, and anyone seeking to view them will need to file a Freedom of Information Act request.
The cryptocurrency market also experienced volatility. Bitcoin saw a slight increase to $65,900 per coin, but overall, the original cryptocurrency has lost significant value, falling from a high of approximately $125,000 per coin in October 2025 to a low of $61,300, according to Fortune. Shares in Michael Saylor's Bitcoin treasury company, Strategy, fell 17% and are down 75% from their peak last year.
In the realm of artificial intelligence, StrongDM's AI team shared details of their work in Software Factories and the Agentic Moment, where coding agents produce code without human review, as reported by Hacker News. The team built a "Software Factory" where specifications and scenarios drive agents to write code, run harnesses, and converge without human review.
Additionally, a paper titled "First Proof" was submitted to arXiv, exploring the ability of current AI systems to answer research-level mathematics questions, according to Hacker News. The authors shared a set of ten math questions that had arisen in their research process.
Discussion
AI Experts & Community
Be the first to comment