AI can now identify anonymous social media users by stitching together details scattered across the internet…
There was a time when you could pretend to be anyone on the internet and no one would be able to figure out your real identity.
Priya Sharma, a marketing strategist by profession, used her real name on her LinkedIn account so that potential recruiters could find her. Her Instagram handle was priya_lens to showcase her photography skills. Most of her friends followed her on that account. But they didn’t know that she also had another Instagram handle called ChaiAndCipher where she never revealed her identity. That was to talk to strangers on the internet without revealing who she was or where she was from.
This used to work. You could have a burner account or a pseudonym and you could just as easily have a second life.
But now, thanks to Artificial Intelligence, especially Large Language Models (LLMs), that game is as good as up.
How LLMS decode Hidden Accounts
A recent study by The Guardian has shown that LLMs excel at decoding anonymous accounts. Earlier, this used to require time, skills and obsessive digital stalking but now can be automated, scaled and executed with unbelievable precision.
Because at the heart of all this decoding and stalking is something that comes very naturally to AI: pattern recognition.
Long story short, every anonymous post that you make reveals something about you. Location hints, writing style, interests, timelines, even emotional undertones. They might not mean a lot individually but AI thrives on aggregation.
It can cross reference data which is spread across platforms and link anonymous accounts to real world identities. And the scary part is that this isn’t even hacking in the traditional sense of the world. There are no passwords being cracked or systems being breached.
AI is successfully piecing this together based on publicly available data.
From Niche Skil to Mass Capability
What is even more alarming is not that capability itself but the accessibility.
Earlier, to crack something like this, one required open-source intelligence, patience and probably institutional backing as well. AI has collapsed that barrier and turned this specialist skill into a publicly-accessible tool.
It has also made such privacy attacks very cost effective which in turn has drastically altered what we used to previously consider as online privacy.
The Consequences of this Development
Apart from not having online privacy anymore, this development could have far reaching consequences.
- Targeted scams: By prying on your information that you believe to be hidden or anonymous, scamsters can be more convincing when they approach you online as they would already have a lot of your data.
- Political dissidents: Previously, such activists used to hide behind hidden accounts. This is how they could reach out to the masses while maintaining their privacy but now the government with its vast resources can easily unmask them online.
- False identification: A problem with AI is that it can be confidently wrong. So, while doing its pattern recognition, even if it makes a mistake, it will vouch for its findings with utmost confidence which could lead to harassment or reputational damage.
The Paradox of Public Data
The underlying fact about all of this is that none of this data is being stolen: it is available online.
Social media has always incentivised oversharing and under the belief that your account is private or anonymous, a lot of information has been volunteered and shared freely.
So, AI does not need new information, it just uses what is already there. And it is aggressive in its approach.
The Last Word
Most privacy frameworks and policies were designed for an older internet. But the internet has evolved while the frameworks and policies have stayed primitive. Researchers behind The Guardian study have been calling for revised frameworks. The problem is that regulations tend to move slowly while AI continues to move at pace.
The final verdict is that online anonymity is next to impossible. It requires a lot of effort: deliberately reducing cross-platform inconsistency, limiting the amount of crucial information revealed online and understanding how details can lead to patterns.
Because in an AI-shaped internet, unlike us, the machines can see the whole picture.
In case you missed:
- Why You Should Never Reuse Passwords
- Watching, Not Sharing: Why People aren’t posting on Social Media anymore
- Governments move to rein in social media and AI for children
- 14 Things to do Immediately if you Lose your Phone
- How Google’s VaultGemma Pushes the Boundaries of Safe AI
- Meta, YouTube declared Liable for Social Media Addiction
- Global Backlash and Legal Pressure Follow Grok’s Explicit Images
- Is AI marking the End of the Web Traffic Era?
- Mary Meeker’s AI Report: ChatGPT is Growing Faster than Google Search
- India’s 3-Hour Rule Signals a New Era for Social Media Platforms
