cutting corners: A sophisticated spam campaign, which takes advantage of Genai’s large language model, has targeted tens of thousands of websites, which reveals the deep side of LLMS. According to a detailed report by Pentinellabs, the outline behind this operation, Akirabot has been dubbed, has successfully bypassed the spam detection filter, conveyed to AI-borne messages to more than 80,000 websites in just four months.
Akirabot is a python-based structure that mainly exploits the website contact form and live chat widget, targeting small and medium-sized businesses. Its goal is to promote suspected search engine adaptation services under the brand “Akira” and “ServiceDrape”.
Unlike traditional spam tools, which rely on the duplicate templates, Akirabot uses Openai’s chat API to generate unique messages to each targeted website. It crafts individual materials using site-specific details, scrapping with the beautifulsoup, making messages more difficult to detect spam filters.
The modular design of the framework includes advanced captcha bypass mechanisms and network theft techniques. It uses the selenium webdriver to simulate legitimate browsing behavior, as well as injections such as scripts to manipulate browser characteristics such as graphics rendering, fonts installed fonts and system memory profiles.
These amendments allow Akirabot to mimic real user behavior, defeating captcha systems such as Hcaptcha and Recaptcha. Additionally, it depends on proxy services like Smartproxy to diversify traffic sources and avoid IP-based restrictions.
Sentinlabs highlighted the archives dating back on September 2024 that documentation of the development of akirbot. Initially referred to as “Shopbott”, Framework expanded its targeting to platforms usually used by small businesses from Shopify-based websites to Godaddy, Wix, Squarespace, and others.
The graphical user of the bot allows interface operators to monitor success matrix and adjust the settings to target many websites. The logs received by the researchers suggest that Akirabot successfully spailed over 80,000 domains while failing over around 11,000 attempts. Overall, more than 420,000 unique domains were targeted.
The use of AI-related material in spam campaigns marks a significant change in strategy. It highlights dual-use nature of large language models: while they strengthen innovations in automation and communication, they also provide equipment for malicious activity.
Openai immediately responded after being alerted by Pentinellabs, disabled the AKIBOT API key and confirmed its commitment to prevent misuse. “It is against our policies to distribute outputs from our services for spam,” Openai said. “We take misuse seriously and are constantly improving our system to detect misuse.”
Despite this, Sentinlabs have warned that Akirabot operators are likely to continue to refine their techniques as the website hosting providers have strengthened the rescue. It was noted that the dependence of the campaign on the captcha displays high levels of refining and determination on bypassing technologies and proxy rotation.
