Icelandic cybersecurity firm urges using AI against AI threats

Everyone needs to remain alert as increasingly powerful artificial intelligence systems pose potential risks to humanity, Theódór Ragnar Gíslason, founder and chief executive of Defend Iceland, told mbl.is.

Recent warnings from within the AI industry have focused on the possibility that the development of more powerful systems could eventually create a serious threat to humanity. Former employees of OpenAI and Anthropic have said the companies are competing to develop more capable AI, without it being clear how such systems can be kept under control.

Theódór said he shared those concerns, but added that people working in cybersecurity and artificial intelligence were also considering how to defend against the development and the ethical questions surrounding it.

Defend Iceland developed AI Defender, Iceland’s first AI-powered hacker. The system can independently identify and verify previously unknown security vulnerabilities in computer systems by using different AI models whose methods increasingly resemble those of experienced specialists.

AI Defender is a standalone product designed to operate alongside Defend Iceland’s bug bounty platform. The platform is based on cooperation between companies and a community of ethical hackers who are authorised to search for vulnerabilities in corporate systems and report them.

“These latest models, including Astra from OpenAI and Mythos, which Anthropic released recently, can do extraordinary things if they are not subject to some kind of restrictions or responsible oversight. They can find security vulnerabilities in almost anything,” Theódór said.

He said the pace of development had surprised the company and that AI Defender had already identified tens of thousands of vulnerabilities.

Theódór believes part of the response to the dangers posed by increasingly powerful AI is to use the technology against itself, identifying vulnerabilities before they can be exploited. AI could potentially find previously unknown weaknesses and use them to attack critical systems, including utility networks, energy production facilities and hospitals.

“I think we all need to remain alert and be conscious of trying to harness this technology as responsibly as possible, so we can defend ourselves before something happens,” he said.