Trumponomics Will Mythos Ruin or Save the Global Financial System?
23 snips
Apr 21, 2026 
Laura Noonan, Bloomberg finance reporter covering IMF/World Bank reactions; Michael Deng, Bloomberg Economics geoeconomics and AI analyst. They unpack Mythos blindsiding policymakers, how it can chain cyber vulnerabilities into large attacks, its potential as a defensive tool, unequal access risks for smaller banks, and the policy tensions over who should control powerful AI models.
AI Snips
Chapters
Transcript
Episode notes
Mythos Finds Zero-Day Flaws Autonomously
- Mythos autonomously finds large numbers of zero-day vulnerabilities at a speed far beyond prior models.
- Anthropic testers ran non-security staff overnight and returned to discover the model had uncovered many previously unknown exploits across browsers and OSes.
Overnight Test Revealed Hidden Exploits
- Testers without security training ran Mythos overnight and found it returned many vulnerabilities the next morning.
- That real-world experiment highlighted Mythos's capacity to surface hidden exploits even for non-expert users.
Agentic Chains Multiply Attack Paths
- Mythos demonstrates agentic chaining: linking minor exploits into full attack campaigns rather than just generating isolated bugs.
- Anthropic showed Mythos finding 181 Firefox exploit paths versus two from the prior model by combining a few vulnerabilities into many attack chains.