AI Tools Are Generating Massive Attacks for Little Cost

In September, a Chinese-speaking attacker employed multiple AI tools, including open-weight and public models, to target 105 e-commerce sites. Within days, at least 27 companies were compromised, with malicious agents stealing over 600,000 payment card records from two firms and installing credit-card skimmers on five others. This campaign also affected at least one Fortune 500 company, according to Gambit Security, which uncovered the attack.

The average cost for the attackers was just $25.46 per successful target. Sa’ar Elias, co-founder and chief research officer at Gambit Security, noted, “The scale and success of this campaign are remarkable for just one week, especially considering the size of some of the victims. Work that would previously have required a coordinated team of skilled attackers can now be carried out by one person without much prior knowledge.”

This autonomous credit-card and data-stealing campaign is indicative of a broader trend. Agentic AI has accelerated vulnerability research, leading to a significant increase in Common Vulnerabilities and Exposures (CVE) assignments. Attackers are increasingly leveraging AI to automate attacks, exploit systems more rapidly, and even function as fully autonomous hackers.

In a separate study, researchers from AI cyber-defense firm Silverfort tested how quickly autonomous AI could escalate initial access to full system compromise using publicly available versions of Claude Opus. The simulated attacker achieved complete domain control in just eight minutes, costing approximately $12.

Colin Ferris, threat hunting and incident response lead at Silverfort, remarked, “The barrier to entry is falling. What it takes is a basic foothold, not exotic skill, and that’s why this won’t stay confined to sophisticated actors.”

Declining Costs for Attackers

The prices in dark markets reflect the decreasing costs for attackers. According to threat intelligence firm Flashpoint, the average price for initial access has dropped from over $1,400 to $429. Despite a decline in the number of companies paying ransoms—28% in the first half of 2026—ransomware activity has surged by 45%.

Ian Gray, vice president of intelligence at Flashpoint, stated, “Initial access is where we’ve measured the price drop. Elsewhere in the attack chain, we see AI making operations faster and easier to scale.”

Challenges for Cybersecurity Defenses

The current landscape resembles modern battlefields, where inexpensive autonomous attacks, akin to drone strikes, challenge costly defenses. Cyberattackers in the latest campaign utilized the agentic AI system Hermes and two AI penetration-testing frameworks, Strix and Cairn.

Elias emphasized that attackers can now launch sophisticated attacks at low marginal costs while defenders must ensure high reliability across complex environments. He warned that defending against every individual attack could become economically unsustainable.

To enhance defenses, companies need to apply AI to existing security processes and reassess their business operations to protect critical functions. Albert Ziegler, chief information security officer at XBOW, noted that every part of a company connected to the Internet is a potential target.

Improving Cybersecurity Measures

To strengthen defenses, organizations should focus on common vulnerabilities exploited by AI agents, such as overprivileged accounts and reused credentials. Automating access decisions can help speed up responses to threats. Ferris explained that when Silverfort researchers improved controls, the same AI agent that previously took control in minutes encountered 21 dead ends and abandoned the attack.

Ferris stated, “AI didn’t create these weaknesses; it just finds them faster. Controls that stop the attack at the point of access are cheaper than cleanup afterward.”

Gambit Security’s Elias suggested that companies should prioritize investments based on their potential impact on business operations. However, he cautioned that adapting to the economics of AI may require increased budgets, emphasizing that spending should focus on effective protection of business processes.

Source: Dark Reading