LinkedIn recently announced it has verified more than 55 million users for free to combat the spread of misinformation fueled by the rise of artificial intelligence. With deepfakes and AI-generated content making it harder to tell real people from fabricated ones, the platform's Head of Trust and Safety, Oscar Rodriguez, framed the company's philosophy this way:
"We would rather get it wrong, as in potentially not verifying someone, as opposed to verifying someone that then we realized it was a wrong verification."
On its face, that sounds cautious and responsible. But it's worth pausing on what's actually happening behind that sentence.
The trade being made
To hit its verification goals, LinkedIn's other verification method involves having users submit their government IDs with partners such as Clear and Persona. That means real people are handing over real government-issued identification; passports, driver's licenses- to third-party identity vendors. Those vendors don't operate in isolation; they run on their own stacks of subprocessors, cloud providers, fraud-detection tools, and data brokers. It's not unusual for a single identity-verification vendor to disclose 15-20+ subprocessors touching that data at some point in the pipeline.
So the actual exchange looks like this: in order to protect users from synthetic people (AI-generated fakes), platforms are asking real people to surrender their most sensitive, real-world identity documents to a web of third parties most users will never see, audit, or fully understand.
"Getting it wrong" shouldn't be the standard
Rodriguez's quote is honest, and honesty is worth something. But "we'd rather get it wrong than get it right" is a strange place for a company handling government ID data to land. It's an admission that the system is imprecise and an implicit ask for users to accept that imprecision as the cost of participating.
Big tech companies routinely operate this way: ship the imperfect version, collect the data anyway, and treat "we're working on it" as an acceptable long-term state. If a smaller company handled ID verification with this level of built-in error tolerance, it would raise eyebrows. At LinkedIn's scale with over a billion members, a stated goal of 100 million verified users by 2025, that tolerance for "wrong" scales too, along with the amount of sensitive data flowing through third-party pipelines.
The real question
Is the threat of AI-generated misinformation serious? Yes. But the fix shouldn't default to funneling real people's real government IDs through sprawling third-party data ecosystems just because it's the fastest path to a headline number. If verification is going to be the industry's answer to AI fakes, the standard should be precision and minimal data exposure, not "we'd rather be wrong."
We let big tech set these terms because we've grown used to it. That doesn't mean we should stop asking for better.
Source : CNBC