Bridgewater, a major investment firm, has flagged a growing concern: artificial intelligence is now amplifying cyber threats beyond what current defences can handle. Their latest research argues that while defenders are patching more vulnerabilities than ever, AI is letting attackers move even faster, exposing a widening gap in the security landscape.
The report points to recent incidents like Microsoft outages and vishing attacks on financial institutions as early signals of the trend. While not all details are public, Bridgewater’s position is clear: these disruptions are a preview of a larger wave of AI-enabled attacks, and waiting for full attribution before acting is risky.
Attackers Are Beating Patch Cycles
The core problem is that AI is enabling attackers to exploit vulnerabilities in the narrow window after a patch is published but before it is widely adopted. Bridgewater notes that the average time from a public disclosure to active exploitation has dropped from about a month to just 11 days. This is the new window defenders must operate in.
When a vulnerability is patched in open-source software, the underlying issue becomes visible to everyone—including malicious actors. Downstream users must update quickly, but AI is helping attackers automate the hunt for fresh bugs and weaponise them before most users can respond. The number of zero-day exploits has roughly doubled since powerful open-source AI models became available.
Defensive Efforts and Their Limits
Initiatives like Anthropic’s Glasswing and OpenAI’s Daybreak are surfacing more vulnerabilities, but the scale is daunting. The report highlights that defenders must address every critical bug, while attackers only need to find one overlooked flaw. Open-source software, which underpins much of the internet, is often maintained by small teams or volunteers with limited resources to keep up.
Bridgewater also observes that as developers use AI coding tools to write more software, the total attack surface expands. Even if AI-generated code is slightly safer, the sheer volume means more opportunities for mistakes and new vulnerabilities.




