Adversarial AI Is Accelerating Cyberattacks, But Slowing AI Isn’t the Answer
Artificial intelligence is becoming more capable, and that means both defenders and attackers are finding new ways to use it.
I recently read Google Threat Intelligence Group’s report, “From Prompting to Autonomy – The Evolution of Adversarial AI.” What stood out to me most was not simply that cybercriminals are using AI. That was predictable. What surprised me was how much AI can reduce the time required to carry out an attack.
Google reported observing threat actors compromise a cloud resource and then plan, build, and execute an AI-assisted credential-harvesting campaign in less than six hours. The report also describes attackers using AI throughout multiple stages of an attack, including reconnaissance, social engineering, malware development, troubleshooting, and automation.
The Real Problem Is Speed
I think one of the biggest cybersecurity challenges created by AI is speed.
Many of the techniques attackers are using are not completely new. Credential theft, malicious software packages, phishing, and cloud compromise existed long before modern AI systems.
The difference is that AI can make these processes faster and easier to scale.
If an attacker can automate tasks that previously required hours of manual work, defenders have less time to detect and respond. That means security systems will increasingly need automation of their own.
In my view, the future of cybersecurity may become a competition between automated attackers and automated defenders.
We Should Not Stop AI Progress
At the same time, I do not think the misuse of AI is a good reason to slow down AI development itself.
Almost every powerful technology can be used for both productive and harmful purposes. The existence of malicious users does not automatically make the underlying technology harmful.
There will always be people willing to misuse technology. Trying to prevent technological progress because bad actors exist would also prevent many of the benefits that legitimate users, researchers, developers, and cybersecurity professionals can gain from the same technology.
A better approach is to continue developing AI while improving the systems that detect and prevent abuse.
Google describes a similar strategy in the report: continuing AI development while strengthening model safeguards, threat intelligence, and systems designed to disrupt malicious activity.
AI Also Creates New Things to Protect
Another part of the report I found interesting is that AI is not only being used as an attack tool. AI systems themselves are becoming targets.
Google observed attackers attempting to steal proprietary models, source code, prompts, API credentials, and other AI-related assets. This means cybersecurity teams increasingly have to treat AI infrastructure as valuable intellectual property and critical infrastructure.
That changes the cybersecurity landscape. Organizations now have to think about protecting not only databases, accounts, and traditional servers, but also models, training systems, AI credentials, and the software supply chains surrounding AI.
My Takeaway
My main takeaway is that AI probably will not fundamentally eliminate the basic principles of cybersecurity. Attackers will still look for weak credentials, vulnerable software, poorly configured systems, and human mistakes.
What AI changes is the scale and speed at which those weaknesses can be exploited.
For cybersecurity professionals, that makes learning automation, threat detection, cloud security, and AI security increasingly important.
I do not think the solution is to stop AI advancement. The more realistic challenge is making sure defensive technology advances just as quickly.
Source: Google Threat Intelligence Group — GTIG AI Threat Tracker: From Prompting to Autonomy – The Evolution of Adversarial AI, September 8, 2026.