Why Anthropic's Escaped AI Agents Still Think They're in a Simulation: DEX in the City

Description

Anthropic's AI agents escaped a hacking test and still think they're inside it. Katherine, Jessi, and Vy Le on who's liable when a model breaks free, plus the $100M Coldcard hack and Kalshi's court losing streak. ========================================================================= 🔥 EPISODE LINK 🔥 https://unchainedcrypto.com/dex-in-the-city/dex-in-the-city-how-claudes-red-teaming-agents-escaped-a-test-without-realizing-it ========================================================================= Thank you to our sponsor: 👉 Cape: Your biggest crypto vulnerability isn't your wallet, it's your phone number. Cape is America's privacy-first mobile carrier that rotates your SIM identity daily and blocks SIM swaps before they happen. Get 33% off your first six months at https://cape.co/unchained (use code: UNCHAINED). ========================================================================= Anthropic's AI models broke out of a fake hacking simulation this week, and some still think they're inside it. One agent invented an email address and phone number to pose as a person, then published malware that twelve companies downloaded before anyone caught it. Katherine Kirkpatrick Bos, Jessi Brooks, and Vy Le use the incident to revisit a theme running through the whole episode: who has a duty to disclose when something breaks, and why crypto and AI are both being left to police themselves. They start with the Coldcard hardware wallet hack, where a firmware flaw cut seed phrase randomness roughly in half, letting attackers brute-force wallets meant to be unguessable. From there, the hosts turn to Kalshi's losing streak in New York courts and the race among builders to acquire a CFTC-registered designated contract market, before landing on Anthropic's own agents slipping past the guardrails meant to contain them. Banks have 36 hours to disclose a breach. AI labs and wallet makers, the hosts argue, are still working entirely on the honor system. ========================================================================= Host: ⭐️ Katherine Kirkpatrick Bos, General Counsel, Host of DEX in the City and General Counsel of Chainlink - https://x.com/kkirkbos ⭐️ Jessi Brooks, General Counsel at Ribbit Capital - https://www.linkedin.com/in/jessi-brooks-8289ab32/ ⭐️ Vy Le - Co-host of DEX in the City and General Counsel of Veda - https://x.com/TuongvyLe12 Timestamps: 🩺 02:16 Katherine Kirkpatrick Bos on joining Chainlink Labs 🔐 03:08 How a Coldcard firmware flaw let attackers guess seed phrases with AI 💙 18:32 Cape: Get 33% off six months of privacy-first mobile service at https://cape.co/unchained ⚖️ 19:36 Why Kalshi keeps losing its fight against New York's gambling regulators 🏛️ 30:47 DCM: the CFTC license every prediction market and perps exchange needs 🤖 33:37 Why Anthropic's AI agents escaped a test and still think they're inside it 💧 45:40 Matt Damon's crypto.com ad money and the water.org donation ========================================================================= #bitcoin #ethereum #crypto #cryptocurrencies ========================================================================= Subscribe to our channel! CHECK OUT THE CRYPTOPIANS → https://www.youtube.com/watch?v=DFzunCFXrv4 WEBSITE → https://unchainedcrypto.com UNCHAINED DAILY NEWSLETTER → https://unchainedcrypto.beehiiv.com/subscribe/ LAURA'S TWITTER → https://x.com/laurashin UNCHAINED TWITTER → https://x.com/unchained_pod FACEBOOK → https://www.facebook.com/unchainedwithlaurashin INSTAGRAM → https://www.instagram.com/laurashin/?hl=en LINKEDIN → https://www.linkedin.com/in/laurashin/ MERCHANDISE → https://unchainedpodcast.company.site/ ++++++++++++++++++++++++++++++++++++++++++++++++ NEW EPISODES OF UNCHAINED EVERY TUESDAY AND FRIDAY! ALONG WITH BITS & BIPS AND THE CHOPPING BLOCK Apple → https://podcasts.apple.com/us/podcast/id1123922160?ls=1 Spotify → https://open.spotify.com/show/1cJrrfGY1SKBIRn5noKSAf?si=6SI4qIcRTEO7EhOe0V9n4w RSS Feed → https://feeds.megaphone.fm/LSHML4761942757 YOUR NO-HYPE RESOURCE FOR ALL THINGS CRYPTO!