Google says artificial intelligence has helped its security teams fix more vulnerabilities in Chrome in a single month than were patched across the browser’s previous two years of major releases.
The company announced that its internal AI-powered security tools helped identify and fix 1,072 security bugs across two versions of Chrome released in June.
That was more than the 1,036 bugs fixed across the previous 23 versions released over the preceding two years.
The figures provide one of the clearest examples yet of how artificial intelligence is beginning to change the economics of cybersecurity.
Security researchers have always faced a basic problem.
Modern software contains enormous amounts of code, and even experienced engineers cannot manually examine every possible interaction between different parts of a system.
A small mistake can sometimes create a vulnerability that attackers can exploit.
AI changes the scale at which that search can happen.
Google says large language models are increasingly capable of examining code, analysing previous changes and identifying patterns that may indicate security problems.
The result is a growing ability to search for vulnerabilities faster than traditional human-led processes alone.
However, finding more bugs creates another challenge.
If AI helps defenders discover thousands of previously hidden vulnerabilities, software companies may need to increase the speed at which they investigate and fix them.
Google has already had to adapt.
The Chrome security team is using AI throughout different stages of vulnerability discovery, triage and remediation, helping engineers identify which problems are most important and develop fixes more quickly.
The development represents a major shift in how cybersecurity experts have predicted AI would affect software security.
For years, researchers have warned that AI could give attackers the ability to find vulnerabilities at a much larger scale.
An AI system capable of examining code continuously could potentially identify weaknesses faster than traditional attackers.
Google’s announcement suggests the same technology may also give defenders a significant advantage.
The race is therefore becoming increasingly symmetrical.
Attackers may use AI to find weaknesses.
Defenders may use AI to discover and fix them first.
The company says the current increase in discovered bugs is partly the result of AI tools searching through areas of code that may previously have received less attention.
That could mean the number of discoveries eventually changes as more of those weaknesses are addressed.
Google is also investing in longer-term solutions, including efforts to improve the security of Chrome’s underlying code.
Some parts of the browser are being moved towards Rust, a programming language designed to reduce certain categories of memory-related vulnerabilities.
That combination could become increasingly common across the technology industry.
AI can help find problems in existing software.
Safer programming practices can help prevent new problems from appearing in the first place.
For Chrome users, the most important message is relatively simple: keep the browser updated.
The increase in vulnerability discoveries does not necessarily mean Chrome has suddenly become less secure.
In many cases, the security problems are being discovered and fixed before attackers can exploit them.
But the scale of Google’s findings demonstrates how quickly the software security environment is changing.
Artificial intelligence is allowing companies to inspect code at a level that would previously have required significantly more human time.
That could lead to safer software.
It could also lead to a future where software updates become more frequent as AI systems continuously identify new issues.
For cybersecurity teams, the message is becoming increasingly clear.
AI is not just another security tool.
It may become part of the fundamental infrastructure used to discover, understand and defend against software vulnerabilities.
The companies that learn how to use it effectively could have a significant advantage.
