August 17, 2026, (Inside AI) — A new criminal AI service called MessiahGPT is selling ransomware, rootkits, and credential stealers on BreachForums for as little as $8 a month. Security researchers at Trellix exposed the platform, which also gives away 50 free prompts without requiring an account.
The tool runs through a web portal at messiahgpt[.]de and an active Telegram channel. Its creator claims the underlying model was built from scratch without moral boundaries. Unlike commercial assistants, it skips Reinforcement Learning from Human Feedback, omits safety frameworks, and blocks no malicious requests.
This is not a theoretical risk. Trellix confirmed the service produces working payloads that cybercriminals find useful enough to buy. The training corpus allegedly includes dark-web repositories, raw web data, leaked manuals, and unedited guides. Those architectural claims remain unverified, but the output speaks for itself.
The discovery matters because it marks a shift in who can launch sophisticated attacks. MessiahGPT does not rely on advanced engineering. Its real breakthrough is affordability and ease of use. A freemium model borrowed from software-as-a-service playbooks turns cyberattack capabilities into everyday consumer goods.
This fits a pattern. In 2023, WormGPT appeared on underground boards to support business email compromise schemes. FraudGPT followed, selling for up to $1,700 a year to generate malware and phishing sites. DarkBERT has also been co-opted by bad actors for social engineering and malware. Hackers have already proven they will pay for unaligned AI.
Cheap access fuels a phishing arms race
Generative AI lets attackers alter phishing messages, fake websites, and malicious scripts in real time. Every changed instance evades basic security filters that look for signatures or fixed phrasing. Advanced phishing-as-a-service frameworks now add browser fingerprinting, anti-bot mechanisms, rotating CAPTCHAs, and hosting filters to dodge detection.
Law enforcement has sounded alarms. Europol documented more than 120 distinct ransomware strains active during 2025. The FBI reported that business email compromise cost victims $3.046 billion in damages that same year. MessiahGPT lowers the barrier further, putting these tools within reach of low-skill criminals.
Trellix researchers uncovered the platform while tracking illicit AI services on underground forums. The subscription runs on crypto payments, which complicates attribution and takedowns. The free tier acts as a gateway, letting potential buyers test malware generation before committing funds.
The service also highlights a structural gap. Commercial AI vendors invest heavily in safety layers. Criminal developers have no incentive to do the same. They can strip alignment, fine-tune on harmful data, and sell access without legal consequence. MessiahGPT is the latest proof that this underground economy is maturing.
What comes next is unclear. Takedowns of such services are rare and often temporary. New domains and Telegram channels appear quickly. The more pressing question is whether defenders can adapt faster than attackers can subscribe.
