AI is exposing security holes across the internet
AI tools are making it easier to find weaknesses in widely used software, accelerating a race between attackers and defenders. Cloudflare says it released protections for a recent vulnerability, while OpenAI, Amazon and Vercel are backing efforts to find and repair flaws. The response matters because the affected software underpins a large share of websites.
AI is exposing security holes across software used by roughly 70% of websites, and security teams are racing to find and fix flaws before attackers exploit them. Stephanie Cohen, Cloudflare’s chief strategy officer, said the moment feels unsettling, but described AI as a tool defenders can also use.
The change matters well beyond technology companies: weaknesses in widely used software can leave websites vulnerable, while the same AI systems that help uncover flaws can also make it easier to exploit them. Cloudflare says it has already released protections for a recent vulnerability, including for customers on its free service.
AI speeds up the search for software flaws
For years, many software weaknesses went unfixed because repairs took time and money, and hackers often lacked the skills to exploit them. AI changes that balance, according to the Business Insider report: a person who once needed years of training to break into a computer system can now get help from an AI model.
These tools can also search for weaknesses faster than people can. The report says the number of newly published software vulnerabilities has surged, creating a race between those trying to break into systems and those working to protect them.
“The world is a little bit scary right now,” Cohen said at Cloudflare’s annual conference in San Francisco. Her company helps protect about 20% of the internet, according to the report.
Warnings online have gone further. Nikita Bier, an app developer and former head of product at X, posted, “It's time to turn off the internet,” on Wednesday night, the report said. Posts on X and Reddit invoking a “Black October” also circulated survival advice, including stockpiling cash and downloading maps.
Security companies turn AI toward repairs
Cloudflare is using AI’s growing ability to identify vulnerabilities as part of the defensive response. The company released protections for the latest vulnerability to its customers, including those on its free service, according to the report.
OpenAI has launched an initiative called Patch the Planet, working with Cloudflare and other security professionals to find weaknesses in widely used software and help repair them. The effort targets the same broad problem: flaws in software can affect many websites, while fixes need to reach the people responsible for maintaining it.
Tom Occhino, Vercel’s chief product officer, said that the company has been speaking with partners about launching a nonprofit. The proposed group would fund AI agents to search the internet for software flaws and help fix them.
Amazon is also entering the effort. Its Continuum service, available Thursday, uses AI agents to find security holes and propose fixes. Amazon said maintainers of freely available software can use the service to help protect the wider internet.
Human review remains part of the process: people may still need to check and approve repairs before they are applied. That leaves software maintainers with both a faster way to identify problems and a need to assess the proposed fixes.
Cloudflare, OpenAI, Vercel and Amazon are pursuing different approaches, but each initiative described in the report aims to turn AI’s search capabilities toward finding and repairing flaws rather than exploiting them.
