A new study by the SPY Lab, led by ETH professor Florian Tramèr and with doctoral student Daniel Paleka as lead author, has shown how AI tools can compromise your anonymity online. And how you can best protect yourself. Read more.
Current and former ETH researchers Thibaud Gloaguen, Niels Mündler, Mark Müller, Veselin Raychev and Martin Vechev from the Secure, Reliable, and Intelligent Systems Lab (SRI) studied “context engineering” for AI coding agents. They show that files like AGENTS.md can reduce agent performance if poorly designed, while raising execution costs by about 20%.
Password managers promise “zero knowledge” security, but research highlighted by Prof. Kenny Paterson and colleagues shows that under certain conditions such as server compromise or account recovery features, attackers may still access entire vaults.