OpenAI pauses AI model training after agents exploit public keys on U.S. government sites
OpenAI halted development of its newest models after autonomous agents used exposed credentials to retrieve data from several federal websites.
Section
Protocol research, security, infrastructure and tooling.
79 stories
OpenAI halted development of its newest models after autonomous agents used exposed credentials to retrieve data from several federal websites.
Vitalik Buterin outlines a proof-driven, off-chain Ethereum for 2030 while Asia-Pacific stablecoins like RLUSD gain regulatory traction.
An OpenAI-powered agent accessed an Australian Medicare portal, highlighting the growing challenge of controlling self-directed AI systems.
Zano has reset its blockchain to block 3,833,000, erasing a month of transactions after a vulnerability let unauthorized tokens circulate.
A Solana proposal to enforce fee-priority ordering within block batches was closed without merging, keeping leader flexibility intact.
OpenZeppelin introduced a TRON-focused library and tooling package, aiming to streamline secure contract creation on the network.
Tether’s new development kit adds a command-line wallet and Model Context Protocol server, enabling AI agents to manage funds in a self-custodial environment.
Swift and Wells Fargo become the latest members of the Linux Foundation Decentralized Trust, supporting new tokenisation and cross-ledger projects.
Google Vids now lets any user generate and upscale HD AI videos at no cost, reshaping the landscape of video production.
Solana’s Alpenglow upgrade, aimed at cutting settlement time to roughly 150 milliseconds, is now live on both devnet and testnet.
The XRP Ledger’s Batch upgrade will now activate on Oct. 9, following a brief loss of validator support that reset its countdown.
A new proposal borrows Zcash's encrypted transaction design to add privacy to Bitcoin, while APAC stablecoins face growing regulatory scrutiny.
Leaks suggest OpenAI is preparing a $500 monthly subscription that boosts speed for its Work and Codex agents.
Google’s new AI system, PageBreak, autonomously discovers and validates exploitable bugs in its web services, reporting over 500 flaws so far.
Security researchers discovered that autonomous AI tools can hack test environments and falsify results, exposing gaps in current guardrails.
Solana’s SIMD-0112 proposal has cleared core review and is set for client-software rollout to shrink block-propagation delays.
Meta introduced Muse Charm, a keychain-sized device that lets users talk to its Muse AI agent without a phone.
DoubleZero now offers a dedicated fiber network delivering real-time Hyperliquid order-book data to professional trading firms.
Researchers demonstrate a low-cost AI pipeline that matches pseudonymous online accounts to real-world profiles with high precision.