Notes from the AI Governance Center: Agentified cyberattacks

"It's going to get a lot worse before it gets better." This is a phrase that I have read a lot over the past few weeks as an increasing number of news reports detail new AI-enabled cyberattacks. 

From last week's PHP crises to reports of more law firms being hacked and Japan's bevy of cyber incidents, I understand why cybersecurity experts share this warning. Hackers' use of AI agents is quickly becoming the norm. These new capabilities are making it easier to access private information for leverage, shut down critical services or perform any number of fraudulent activities.

What is different about AI-enabled cybersecurity?

As a result of a surprising number of cyber breaches in Japan that have affected a wide array of organizations from ride-sharing services to government offices to financial infrastructure providers, it is undeniable that AI is creating new opportunities for cyber hackers. They are able to do more in less time with fewer people as a result of AI agents taking on many of the tasks. 

Masatoshi Kumagai, CEO of GMO Internet, one of the companies affected by the breach and now in need of compensating customers told The Wall Street Journal, that "the underlying issue is the explosive evolution and misuse of AI technology." He added, "attackers are no longer targeting individual sites; they are attacking the entire internet." 

The comments came after a security team found hackers allegedly misused an AI tool built for penetration testing. Following the breach, the attackers demanded a USD50 million ransom and customers experienced unauthorized transactions.

Similarly, given the ability to work persistently on a problem, AI agents were able to uncover unknown vulnerabilities in internet programming language PHP. Matthew Price, CEO of Cloudflare, said on X, that his network saw thousands of attempts in the 10 hours before the announcement, tens of thousands in the 45 minutes after it, and millions soon afterwards. 

Targeted industries 

It seems that where there is sensitive information these hackers are becoming quick to prey. This includes law firms. According to Reuters, citing CrowdStrike's 2026 Legal Threat Landscape Report, "cyber adversaries named 455 legal entities on websites used to post hacked materials, up 39% from the prior year." The concentration of confidential information within law firms can make them particularly appealing targets for attackers seeking leverage, intelligence or financial gain.

Healthcare has seen an increase in AI cyber incidents within the last couple of weeks. The University of New Mexico Hospital recently canceled all appointments, surgeries and campus classes due to cybersecurity threat targeting its computer system. 

Similarly, in Ohio, patients at University Hospitals in Cleveland couldn't access their digital healthcare information as a result of a cyberattack and the hospital had to reschedule procedures and appointments. Kidney dialysis chain DaVita agreed to pay USD15 million to settle a class-action lawsuit over a 2025 ransomware attack that compromised patient data. 

Enforcement and regulation

The increase in cybersecurity incidents has prompted efforts worldwide to establish safeguards through security requirements, enforcement mechanisms and potential liability frameworks. In the U.S., lawmakers today, pressed leading AI companies for information on security risks as Congress considers how to address gaps in AI governance. In September, the White House and several major AI developers announced a "morally binding" safety commitment that included a self-regulatory framework and a four-point AI safety plan centered on independent audits and robust internal controls. In a recent article, IAPP Cybersecurity Law Center Managing Director James Dempsey examined the "policy and regulatory whirlwind" taking shape as governments and regulators respond to AI-related cybersecurity incidents.

What can you do

According to the experts the best defenses are still the foundational basics:

• Update promptly. Security updates or patches are your best protection when attackers move this fast. Turn on automatic updates for your devices and apps.
• Know what you run. Forgotten systems are easy targets. Organizations should keep a list of their software, including old websites.
• Protect your accounts. Use a unique password for each account and turn on multifactor authentication, which asks for a second proof of identity, such as a code on your phone.
• Reset after a breach. Change passwords and security keys if a service you use is hacked, as China's industry ministry advised during the PHP crisis.
• Have a plan. Know who to call and which services can safely be switched off, as the precautionary shutdowns showed.

AI doesn't change the fundamentals of security. It just makes skipping them far more costly.

Further reading

• An Alignment assessment of cybersecurity incidents, Anthropic 
• Japan Vows to Fortify Cyber Defenses as GMO Hack Hits Forex Brokers, Wall Street Journal
• Cybertips for non-developers, Deana Burke @wetclaude
• What is a cyberattack?, Fortinet