A Magic Eden marketplace exploit resulted in approximately 8,472 NFTs worth ~$1.7M being transferred from 447 collections, including BAYC and Azuki, through abandoned on-chain approvals after the platform shut down its EVM marketplace in March. A Yuga Labs executive claims the operation is a whitehat rescue, though confirmation awaits return of the NFTs to their owners.
A social media discussion about political approval ratings and partisanship, with a user arguing that party loyalty is more stable than approval metrics and questioning the need for a malicious explanation for observed voting patterns.
An FDA panel voted in favor of Grail's Galleri blood test, which screens asymptomatic patients for multiple cancers. The mixed vote represents a key step toward FDA approval for the early cancer detection technology.
A company uses an AI agent to handle routine administrative tasks like email sorting, meeting transcription, and invoice processing, with a human approval gate for all outgoing communications. The system runs on scheduled jobs with strict rules: alerts only for true urgencies, no autonomous sending, and full audit trails of all operations.
An FDA advisory committee will vote today on whether to recommend approval of Galleri, a multi-cancer blood test made by Grail that claims to detect up to 50 cancer types. The decision could lead to insurance coverage for the $949 test and validate the emerging field of blood-based cancer screening, though the test did not meet its primary endpoints in clinical trials.
Loopjacking is a security vulnerability where humans approve one operation but agents execute a different one through representation misrepresentation or post-approval state substitution. Researchers demonstrated the attack across multiple agent platforms including Agno AgentOS and LangGraph, while OpenAI's SDK resisted it through serialized continuity checks.