Artificial intelligence is changing cyberattacks from the work of individual hackers into coordinated operations potentially involving hundreds of autonomous agents, creating a speed problem that traditional human-led cybersecurity systems may struggle to match, according to AI cybersecurity expert Yagub Rahimov.
Rahimov, founder and CEO of Polygraf AI, told FOX Business that the emergence of so-called “AI swarms” doesn’t necessarily mean a cyber doomsday is imminent, but said organizations need defenses capable of responding as quickly as AI can operate.
“You cannot secure machine-speed activity with human-speed monitoring,” Rahimov said. “Machine-speed risk requires machine-speed security, and that’s where we need [to] focus.”
But what exactly is an AI swarm? Rahimov described the progression as beginning with a bot and ending with a swarm.
BANKS WARN AI SHOPPING AGENTS COULD INCREASE RISK OF SCAMS, FRAUD AND DATA PRIVACY BREACHES
A traditional bot performs a narrow task, Rahimov said. A chatbot responds to prompts, while an AI agent can independently carry out a task. A swarm, he said, involves multiple AI agents communicating, dividing up work and coordinating toward the same objective.
“Swarms are coordination of that digital worker… I would call it, in simple wording, a digital workforce,” he said. “They coordinate, they talk to each other and they act as a group to deliver a mission that they’re on.”
In recent months, major AI companies have disclosed incidents involving AI agents that circumvented controls or reached beyond their intended testing environments.
OpenAI previously disclosed an incident in which models undergoing an internal cybersecurity evaluation circumvented controls designed to isolate them from the internet and accessed systems belonging to Hugging Face.
AUSTRALIAN PRIME MINISTER SAYS OPENAI AGENT ACCESSED GOVERNMENT HEALTH WEBSITE, RAISES ‘EXTREME CONCERN’
More recently, Google’s Gemini accessed protected systems belonging to three real companies during a cybersecurity test after internet access was unintentionally made available.
The incidents gained widespread attention and fueled concerns about increasingly advanced AI systems.
Rahimov, however, said there was no reason to panic and pushed back on what he viewed as alarmist reactions to the incidents.
“Most of these news things are not as dangerous as what the news is trying to demonstrate,” he said.
“Yes, it can be, you know, in wrong hands, it can be bad,” he continued. “But for that reason, we have the guardrails. We have the security measures and so on.”
Rahimov argued that as potential AI threats become more advanced, security tools are also improving.
GOOGLE GEMINI ACCESSED PROTECTED SYSTEMS OF 3 REAL COMPANIES DURING ARTIFICIAL INTELLIGENCE CYBERSECURITY TEST
What’s changing with AI, Rahimov said, is that the technology can increasingly find a vulnerability, understand what it means and decide how to act on it without waiting for a human analyst.
“The speed is changing, the ability to comprehend is changing,” he said.
“Instead of finding that failure and bringing it to [a] human and the human analyzing and taking an action… we are seeing [the] machine making sense of it too and, at a much faster scale, making a decision on what to do.”
One of the strongest defenses for organizations, Rahimov said, is controlling what information an AI agent can access and, perhaps most importantly, what actions it has authority to execute.
He added that autonomy should only be expanded after systems demonstrate they can be trusted.
Despite the risks, Rahimov expressed optimism about the future of AI security.
CLICK HERE TO GET FOX BUSINESS ON THE GO
“The future of AI security is controlling… what AI can see, what AI can share, what AI can do,” he said.
“Companies and people shouldn’t have to choose between adopting AI and controlling their security or data.”
Read the full article here