AI is getting out of hands and the pace of its unhinged activities is increasing at such a level that we almost see a case or two every alternate day. Now, in a recent study, the researchers have found that generative AI has made it pretty easy for hackers to reveal the real identities of anonymous social media users. As per a report by The Guardian, large language models can now successfully identify anonymous online profiles and share the actual identities by synthesising the harmless information available on different platforms.
AI researchers Daniel Paleka and Simon Lermen found out that the tech behind platforms like ChatGPT brings down the cost of sophisticated privacy attacks on a dramatic level. The researchers ran an experiment where they gave anonymous accounts into an AI and instructed it to scrape all the information available online.
AI Gone Rogue? 5 Major Chatbots Reportedly Found Promoting Illegal Gambling Sites
In a hypothetical scenario, an anonymous user mentioned his struggles at school and walking their dog, Biscuit, via ‘Dolores Park.’ After that, the AI model ran searches on different platforms for the mentioned details and successfully matched the same to a real world identity with a high degree of confidence.
Moreover, the researchers also warned that the expertise needed to trigger sophisticated attacks is now much lower, with hackers only requiring an internet connection and access to publicly available language models. However, the process has its own drawbacks; for example, in some cases, the potential matches could be too high in number to sort out the exact one.
These types of capabilities are exactly what anthropic is scared of and why the company broke its deal with the US Department of Defense. This kind of practice could be used to surveil citizens of a country if the government has access to it. And yes, the same could also be used to trigger highly personalised scams as well. On a concluding note, the readers are suggested to limit and restrict what they post on different platforms and be aware of any scam if they are approached from the front.
