Ethereum Foundation Reveals Critical Node Vulnerability Found via AI, Highlights Irreplaceable Role of Human Review

The Ethereum Foundation security team used AI agents to test validator node software, uncovering a remote-triggerable vulnerability in the gossipsub protocol that could crash nodes. Researchers emphasized that human review remains crucial to distinguish real threats from AI “hallucinations” among voluminous automated reports. The vulnerability has been patched.

Read More

JPMorgan Tests AI Portfolio Manager: Outperforms 60/40 Strategy in Backtests, But Warns of Crowding Risks

JPMorgan is testing AI agents that autonomously adjust stock-bond allocations. In 20-year backtests, the top model outperformed the classic 60/40 portfolio by 0.7% annually with lower volatility. While all eight tested agents showed better risk-adjusted returns, the bank cautioned results are simulated and widespread AI adoption could lead to strategy crowding and amplified market swings.

Read More