Criminal AI tool Kriminal is mostly just Grok with a jailbreak, ThreatDown finds

What happened

ThreatDown, the security division of Malwarebytes Inc., exposed that Kriminal, a rising criminal AI tool, is essentially just SpaceXAI’s Grok model operating with a jailbreak. Despite Kriminal’s claims of circumventing legitimate AI services, it relies on rented access to Grok, a legal and commercial AI product. This discovery comes as Kriminal gains traction in illicit circles, selling AI capabilities that it does not independently own or develop.

Why it matters

This finding punctures the illusion that criminal AI tools always operate on custom-built clandestine models. Instead, some leverage legitimate AI services through unauthorized hacks or jailbroken access to bypass controls. For businesses and security operators, this raises the stakes for monitoring how AI models are accessed and used, increasing the risk that compromised commercial AI platforms could fuel illegal activities. It also pressures legitimate AI providers to tighten access controls and deepen model security to prevent misuse.

What to watch next

Expect closer scrutiny on AI service providers’ security measures and how easily jailbreaking can enable criminal use. Investigations into whether other illicit AI tools similarly piggyback on commercial models could expand. For defenders, tracking the evolution of jailbroken AI technologies will be critical to anticipating new threats and plugging security gaps that let adversaries repurpose legitimate AI resources.

AI Quick Briefs Editorial Desk