Hadrian unveils AI agents to pre-empt cyber attacks

Cybersecurity firm Hadrian has launched a new version of its agentic artificial intelligence platform that uses autonomous "offensive" software agents to search for and test security weaknesses before criminal hackers reach them.

The launch comes as security researchers report a sharp increase in the use of AI tools by attackers, and as the UK emerges as a major target for cyber incidents in Europe.

Hadrian said its platform runs continuous tests against organisations' external digital infrastructure. The system imitates real-world attack techniques and attempts to exploit weaknesses that it discovers.

The company positions the technology as a response to a growing "machine versus machine" dynamic in cyber security, where both attackers and defenders are deploying automated agents.

Recent industry research cited by the company indicates that the UK accounts for around 25% of all reported cyber incidents across Europe. Organisations in the UK faced more than 100 million malicious attack attempts in a single quarter. The country ranked as the third most targeted globally, behind the US and Canada.

Attackers adopt AI

Security teams have reported broader use of AI by criminal groups over the past year. The tools range from basic chatbots that support fraud attempts to customised models that generate malicious code on demand.

Hadrian pointed to examples such as ransomware gangs experimenting with AI chatbots and "ScamAgents" that run convincing fraudulent conversations with victims. Other cases involve so-called jailbroken models, such as WormGPT, which generate malware at scale.

Attackers also make use of widely available large language models. These include systems such as Anthropic's Claude, which can support tasks from reconnaissance to social engineering when placed in the hands of malicious users.

The company said this trend has increased both the speed and the volume of attempted intrusions, and has lowered the barrier to entry for less skilled attackers.

Rogier Fischer, Chief Executive of Hadrian, said AI agents now sit at the centre of a new phase in cyber conflict.

"We are witnessing a cyber arms race between AI," said Rogier Fischer, CEO Hadrian. "This is no longer a human-versus-human contest, it's machine versus machine. Hackers are building autonomous attackers, we build autonomous hackers on the side of the good. Our AI agents strike first: they expose the weaknesses before attacks have the chance to exploit them and break in."

Autonomous agents

Hadrian describes its platform as a set of autonomous offensive agents that operate on a continuous basis. The agents simulate live attacks against external-facing systems and digital infrastructure.

The platform also ingests information about emerging AI-driven attack patterns. The software uses that data to adjust its testing routines and to mirror the behaviour of current threat actors more closely.

Hadrian said the system can support autonomous actions that expose vulnerabilities before criminal groups discover them. The agents repeat these tests at regular intervals.

The company said this model offers a more persistent and proactive style of defence than traditional, infrequent penetration tests that rely heavily on manual work by human consultants.

The launch underscores a wider shift in the security market towards tools that incorporate automation and AI. Vendors are investing in products that scan corporate networks and internet-facing systems on a continual cycle and generate remediation advice for internal IT teams.

Hadrian said organisations can no longer rely solely on human analysts in an environment where attackers use AI systems to attack at greater speed and scale.

Industry context

The research referenced by Hadrian is based on a review of recent reports and news coverage that document increased use of AI in cyber attacks. The findings align with trends tracked by multiple industry bodies and security vendors, which have warned that AI systems are increasingly present at each stage of the attack chain.

Security consultants report that AI tools now support tasks such as scanning for exposed systems, crafting phishing emails with few language errors, impersonating voices and faces, and writing or modifying exploit code.

Law enforcement agencies and regulators have flagged the risk that this could overwhelm organisations that do not upgrade their defences. The concern grows in sectors that operate legacy systems or hold large volumes of sensitive data.

Hadrian focuses on offensive security testing from what it describes as a "hacker's perspective". The company deploys AI technologies that attempt to identify and mitigate vulnerabilities before criminal actors can exploit them.

The firm offers continuous monitoring of customers' external attack surfaces and carries out ongoing threat analysis. It works with companies in multiple sectors that operate complex digital infrastructures.

Hadrian plans to expand its use of autonomous agents as attackers increase their reliance on AI-driven techniques and tools.