particularly at risk

Last month, Anthropic revealed its new artificial intelligence model, Mythos Preview, which had uncovered thousands of vulnerabilities in major operating systems and web browsers. However, access to Mythos is limited to about 40 tech firms and institutions, leaving many others, including central banks and governments, vulnerable to cyberattacks.

Cyberattacks have surged worldwide, with AI-enabled entities increasing attacks by 89% in 2025 compared to the previous year. AI systems, such as Mythos, are capable of finding and exploiting software vulnerabilities more efficiently than humans, making it easier for criminals and foreign actors to target individuals, companies, and national infrastructure.

The rise of AI in cybersecurity presents a challenge for organizations lacking resources, as widely used software tends to be patched quickly, while proprietary offerings may struggle to address vulnerabilities in a timely manner. For instance, a North Korean hacker group utilized AI tools from OpenAI and Cursor to steal up to $12 million in cryptocurrency from victims.

Since the launch of OpenAI’s ChatGPT, criminals have adopted AI tools to create phishing emails, deepfake videos, voice clones, and malware. Additionally, AI can reverse-engineer security patches, identify vulnerabilities, and generate weaponized exploits within minutes.

When a software firm releases a security patch, AI can analyze it, pinpoint the vulnerability it rectifies, and develop a functioning exploit in a matter of minutes. Security "has to work for everyone, not just at the top of the market," according to Chandramouli Dorai, chief evangelist of cyber solutions at Zoho Corporation. However, a 2024 cybercrime index revealed that Russia, Ukraine, China, the U.S., and Nigeria are among the top sources of cybercrimes, including malware coding, ransomware, data theft, and scams.

Americans lost at least $10 billion to cyber scams originating in Southeast Asia in 2024, with scam centers in Myanmar and Cambodia using AI to find more victims, primarily in the U.S. and Europe.

The recent cyberattacks linked to Iran following Israel and the U.S. attacks on Iran have impacted companies across the Middle East. Moreover, an Iran-linked group claimed responsibility for a cyberattack on American medical technology company Stryker on March 11, causing disruptions worldwide. Even Mythos is vulnerable, as a group of Discord users gained unauthorized access to the model.

OpenAI has responded to Mythos, while Chinese AI companies are developing similar models. Given the rapid advancements in AI, these capabilities are likely to proliferate, potentially beyond actors committed to using them responsibly, according to Anthropic. The consequences for economies, public safety, and national security could be severe.