OpenAI and Anthropic Models Can Easily Hack Websites, Raising Massive Cybersecurity Concerns

SpencrGreenberg · x · 2026-08-01

The author points out that OpenAI and Anthropic currently possess models capable of hacking many websites and companies with ease, while open-source models are only 4-12 months behind in such capabilities. This raises significant security concerns: bad actors or governments could soon leverage these powerful models to launch massive cyberattacks against enterprises.

Related event: Frontier AI Models Show Powerful Hacking Capabilities, Raising Security Concerns(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →