Science & Tech · Cybersecurity · 8 hrs ago
AI agents automate vulnerability discovery and cyberattacks
Google's Threat Intelligence Group found an AI framework that can find software weaknesses and write attack code without step-by-step instructions from a person.
The Russian-language tool, called Exploit Agent, compares software code before and after a security patch to identify flaws.
That could help attackers target internet-connected systems that have not yet installed available patches.
Other AI-enabled threats include stealing access to AI models and using AI to produce large amounts of fake content.
In a survey of Asia-Pacific business leaders, half said their security tools were not ready for AI, while F5 said companies should combine monitoring and automated tests with protections that can block attacks before patches are installed.
Gartner expects global AI spending to rise 47.5% next year to about $4 trillion, with most of that spending going to infrastructure.
Google’s Threat Intelligence Group reported an AI framework that found software vulnerabilities and automatically developed attack code.
The Russian-language framework, called Exploit Agent, compared code before and after security patches.
The report said linked AI agents can carry out attacks with less human involvement.
A survey found that 50% of Asia-Pacific business executives considered existing security solutions inadequate for AI threats.
Gartner forecast global AI spending of about $4 trillion next year, up 47.5% from this year.
- Who
- Google’s Threat Intelligence Group reported the case. The framework was called Exploit Agent.
- What
- An AI framework found vulnerabilities by comparing software code before and after patches, then automatically developed attack code.
- When
- The report was published on November 22, 2026.
- Where
- Not stated
- Why
- Attackers can use the framework to identify and exploit vulnerabilities in systems that have not yet applied security patches.
This story does not have two clearly opposing sides.
This story does not have a timeline yet.
- Report date
- November 22, 2026
- AI framework
- Exploit Agent
- Survey finding
- 50% of Asia-Pacific business executives said existing security solutions lacked sufficient AI response capabilities
- AI spending forecast
- About $4 trillion globally next year, up 47.5% from this year
- AI infrastructure forecast
- About $2.4 trillion, or 59% of next year’s AI spending
What people are saying
Demand from users seeking to scale inference workloads is growing faster than expected, and training workloads are also requiring more and more computing resources