Hackers have also learned a new advanced approach to taking advantage of X chatbot on Grok AI, exploiting the security controls of this chatbot and circumventing the posting limitations to send harmful content to millions of users using a technique researchers refer to as “Grokking,” which allows dangerous malware distribution campaigns to be promoted by the trusted status of the AI assistant.
Threat actors bypass X security measures
BleepingComputer reports that threat actors are taking advantage of Grok, the in-built AI assistant of X, to circumvent limitations on link posting that the platform added to curb malicious advertising. Mavertisers, as found by the researcher of the Guardio Labs, Nati Tal, will tend to have sketchy video advertisements with adult bait content and no link to the proper body, so they are not blocked by X.
They instead put it in a small (From:) metadata field beneath the video card, which seems not to be scanned by the social media site to detect malicious links. X security systems do not normally consider this field, and it presents a weakness through which cybercriminals can utilize it.
Then, probably the same actors pose to Grok through a response to the ad, such as, Where is this video, or What is the connection to this video. Grok decodes the malicious buried field, From:, and responds with the entire malicious URL in a clickable format, which users may then choose to click and be redirected to the malicious site.
The credibility that Grok has increases the malicious content
Since Grok is an established system account on the X platform automatically, there is a boost in credibility, reach, SEO, and reputation of the link, which has the probability of being shared with many users. The researcher has discovered that most of these connections lead to shady ad networks, which result in scams, like fake CAPTCHA tests, information-stealing malware, and other bad payloads.
Rather than being blocked by X, they are rather advertised to users on the platform using malicious advertisements that are strengthened even more by Grok. Tal refers to this trick of taking advantage of this loophole as Grokking and that it is highly effective, in other cases multiplying malicious advertisements to millions of impressions.
Technique finds millions of users
The “Grokking” method is extremely useful in going around the security systems of X. In reported instances, malicious ads have hit millions of impressions using this technique, which illustrates the magnitude of the potential damage that can be caused by taking advantage of AI assistant vulnerabilities.
Scanning all fields, blocking hidden links, and providing context sanitization to Grok (so the AI assistant does not echo links without thinking when requested by users) are but a few of the possible remedies. Such actions would serve to avoid the unwarranted tendency of the AI to facilitate the spread of harmful information.
Security researcher reports vulnerability
Tal attested to making contact with X to report the problem and unofficial confirmation that the Grok engineers were notified of the report. BleepingComputer also called X to inquire whether they know about this abuse and whether they are going to take action about it, but had not replied by the time of publication. The use of Grok Points to more general issues with AI assistants being used as tools to tamper with matters of security and how cybercriminals manipulate trusted automated systems.
Grok exploitation method is an alarming development in the methods of cybercrime since it reveals the ability of threat actors to use an AI assistant to circumvent the safeguards and access huge numbers of people. This weakness justifies the urgent necessity of inclusive AI security measures and real-time responses to the platform in the face of new threats.