Cybersecurity

Anthropic and OpenAI launched their most advanced models to date, focusing on autonomous computer use and 1-million-token context windows. These same models can now execute multistage cyberattacks, mimicking breaches like Equifax using only standard tools. While corporations deploy AI coworkers to automate six-week tasks in a single day, synthetic social networks like Moltbook have already exposed millions of user API keys. Key Takeaways: • AI agents can now autonomously execute major data breaches: Claude Sonnet 4.5 successfully imitated the Equifax breach in simulations by identifying and exploiting known vulnerabilities using only standard open-source tools. • Security failures in synthetic social networks are exposing millions of corporate credentials: Cybersecurity firm Wiz discovered a database breach in Moltbook that exposed 1.5 million API keys used by autonomous agents. • Stablecoins are entering the European mainstream ahead of MiCA deadlines: Fipto’s dual licensing allows it to manage the entire payment chain under a single compliance framework, facilitating near-instant settlements for treasury and merchant payouts.