Anthropic Warns of AI-Driven Cybersecurity Threats to Critical Infrastructure

By Staff, Agencies

A senior cybersecurity official at US artificial intelligence company Anthropic has warned that a sharp increase in newly discovered software vulnerabilities could overwhelm existing defenses in the coming months, exposing critical infrastructure, including electricity and water systems, to heightened cyber threats.

Jason Clinton, Anthropic’s deputy chief information security officer, described the looming challenge as a “vulnerability tsunami,” warning that organizations responsible for essential services could struggle to repair security flaws quickly enough as artificial intelligence accelerates their discovery.

Clinton urged organizations to isolate sensitive computer networks from external access wherever possible, making it more difficult for hackers to penetrate systems that support critical operations.

“As bad as things have gotten in the past 4 days, they are going to get a lot worse before they get better,” Clinton wrote on X.

He also called on security leaders to seek immediate support from their organizations’ boards, emphasizing the need to strengthen basic cybersecurity measures rather than focusing exclusively on AI-related initiatives.

Clinton warned that organizations should prepare for approximately 18 months of heightened cybersecurity risks as the pace of vulnerability discovery continues to increase.

His remarks come amid growing concerns that AI tools could enable security researchers and malicious actors alike to identify software weaknesses faster than companies and government agencies can address them, potentially creating more opportunities for cyberattacks.

On Wednesday, developers of PHP, a widely used programming language that powers websites and online services, released emergency security updates to address two critical vulnerabilities. The flaws could allow attackers to execute malicious code on affected servers without requiring login credentials.

Both vulnerabilities received severity scores of 9.2 out of 10, prompting developers to urge organizations to install the updates immediately.

Clinton did not specify which recent incidents prompted his warning and did not claim that the PHP vulnerabilities were connected to the developments he referenced.

Anthropic has previously reported using its AI models to identify thousands of vulnerabilities in publicly available software, underscoring the technology’s growing capacity to uncover security weaknesses.

The warning highlights a widening challenge for cybersecurity teams: as AI accelerates the discovery of flaws, organizations may face increasing pressure to update systems, strengthen network protections and address longstanding security weaknesses before attackers can exploit them.