The attacks took place with almost no human intervention: intervention was only necessary four to six times per target. The systems carried out most of the activities independently – automated, scalable and precise. A team of people would have needed significantly more time to do this.

Anthropic, whose technologies were used for the major attack, describes the scale as a “sheer volume of work” made possible by AI.

The attack makes it clear that automated attack operations are no longer a future scenario. They are already an operational reality in which attackers use high-performance AI to carry out tasks that would previously have required enormous human resources.

Trust becomes the crucial currency

David Colwell, Vice President of Artificial Intelligence and Machine Learning at Tricentis, sees this as an unmistakable warning signal. In the age of agentic AI, trust will become the hardest and most important currency in business.

The same technology that identifies customers or creates personalized communications in marketing can be used just as effectively for social engineering. Tools that analyze web content or create company profiles can be used for targeted phishing campaigns without technical hurdles.

The crucial weak point: Autonomous AI systems have no intention. They follow instructions – without moral or safety-related evaluation.

Making risks manageable: AI as an unreliable communication partner

Colwell is therefore calling for a fundamental rethink. In the future, companies will have to treat AI requests like automated web crawlers: as fundamentally untrustworthy entities that potentially look for security gaps.

This brings a central factor into focus: transparency. An AI system whose activities cannot be tracked cannot be controlled – and therefore not safe. “Trust comes from clear traceability, verifiable processes and precise governance rules. Without these foundations, every interaction with autonomous systems remains a risk,” says Colwell.

Tricentis: Quality and safety assurance in times of autonomous AI

Tricentis is one of the world’s important players in the area of ​​continuous testing and quality engineering. The completely codeless, AI-powered testing portfolio enables continuous and automated quality assurance – an approach that is becoming increasingly important in the age of autonomous attack systems.

The more autonomous software and AI act, the more important it becomes to ensure quality, security and governance from the ground up. This is the only way companies can use the potential of agentic AI without being at the mercy of its arbitrariness.