AI Is Now Helping Security Teams Find and Fix Vulnerabilities Faster

Artificial intelligence is changing how quickly security teams can discover and respond to vulnerabilities. As attackers increasingly use AI to identify weaknesses and build attack paths, cybersecurity companies are also using the technology to continuously test systems and identify exposures.

Palo Alto Networks has introduced Unit 42 Continuous Frontier AI Defence, an agentic security service designed to discover vulnerabilities, validate their exploitability and help organisations remediate them.

AI Is Changing the Speed of Vulnerability Discovery

Palo Alto Networks says threat actors are increasingly using agentic AI to discover, validate and chain vulnerabilities at machine speed. In one recent Unit 42 investigation, an AI-assisted attack used more than 50 MITRE ATT&CK techniques and reduced weeks of intrusion activity to less than 10 hours.

The company also says automated adversary scanners can weaponise newly disclosed CVEs within 15 minutes.

Unit 42 says the AI-assisted attack reduced weeks of intrusion activity to less than 10 hours, making the speed of AI-driven attacks a growing concern for security teams.

As security testing moves toward faster and more continuous vulnerability discovery, techniques traditionally associated with ethical hacking are becoming increasingly relevant to understanding how AI-assisted attacks can identify and exploit weaknesses.

Why Unit 42 Is Using Multiple AI Models for Security Testing?

Unit 42 Continuous Frontier AI Defence uses multiple AI models rather than depending on a single model for security testing.

Palo Alto Networks says its research found that:

• No single AI model identified more than 40% of vulnerabilities in a complex environment.

• Leading cyber models had less than 10% overlap in the exposures they identified.

• A multi-model approach can help address the blind spots of individual models.

No single AI model identified more than 40% of vulnerabilities in Unit 42’s evaluation, highlighting why the company uses multiple models for security testing.

As AI becomes part of security testing, developers also need to consider how vulnerabilities can emerge across applications and codebases. This makes a strong foundation in software development with AI increasingly relevant when building and securing modern applications.

Finding a Vulnerability Is Only the First Step

Finding a vulnerability does not necessarily mean that an attacker can exploit it. Unit 42 is designed to test whether weaknesses can actually be exploited and whether several vulnerabilities can be chained into a larger attack path.

The service can assess:

• Web applications and APIs

• Cloud infrastructure

• Source-code repositories

• Network assets

It can also provide prioritised remediation recommendations, code-level guidance and virtual patch options.

The service is designed to move beyond vulnerability discovery by validating end-to-end attack paths and providing prioritised remediation guidance.

Many Exploitable Exposures May Not Have a Known CVE

Palo Alto Networks says that, across third-party applications in its customer engagements, two out of three validated exposures had no known CVE. The company also reports that 37% of identified exposures were rated high or critical in severity.

This means security teams cannot always rely only on databases of known vulnerabilities. Testing also needs to examine how weaknesses interact within a specific environment.

Two out of three validated exposures in third-party applications had no known CVE, according to Palo Alto Networks’ customer engagement data.

Unit 42 Moves Vulnerability Testing Toward Continuous Security Checks

Unit 42 Continuous Frontier AI Defence is designed to continuously test an organisation’s environment rather than relying only on periodic assessments.

The approach combines:

• Continuous vulnerability discovery

• Multi-model AI security testing

• Adversary simulation

• Attack-path analysis

• Remediation guidance

During Palo Alto Networks’ internal deployment, the company says continuous Mythos-based scanning achieved more than a year’s worth of traditional penetration-testing results in three weeks. It also identified 3.2 times more high- and critical-severity vulnerabilities per product and helped engineering teams reduce mean time to remediate by 51%.

Palo Alto Networks reports that its internal deployment produced more than a year’s worth of traditional penetration-testing results in just three weeks.

AI Is Moving Deeper Into the Cybersecurity Workflow

The launch of Continuous Frontier AI Defence shows how AI is moving beyond security analysis into vulnerability discovery, attack simulation, exploit validation and remediation.

For professionals working at the intersection of AI and cybersecurity, understanding how these systems are built and applied is becoming increasingly relevant. This also connects with the growing importance of AI engineering in developing and deploying AI-powered applications.

AI-powered security is moving vulnerability testing from periodic checks toward continuous discovery and faster remediation. Palo Alto Networks’ Unit 42 shows how agentic AI can help security teams identify and address exposures at machine speed.