Meta’s decision to remove end-to-end encryption from Instagram direct messages, effective May 8, 2026, is being viewed by some analysts in the context of the company’s expanding artificial intelligence ambitions. The change was disclosed through a quiet help page update. The timing has raised questions about whether AI development is a factor in the encryption decision.
Encryption on Instagram was introduced in 2023 as an opt-in option following Zuckerberg’s 2019 commitment. The feature attracted few users, giving Meta its official justification for removal. However, analysts note that encrypted messages cannot be used to train AI models, while unencrypted ones can.
After May 8, Meta will have access to all Instagram DM content. The volume of conversational data this represents is substantial. For a company investing heavily in large language models and AI assistants, access to real-world conversation data is potentially very valuable.
Tom Sulston of Digital Rights Watch raised the commercial dimension of the decision explicitly. He argued that while Meta may not immediately use DM content for AI training, the commercial pressure to do so will be enormous. Law enforcement agencies including the FBI, Interpol, and national bodies in Australia and the UK had also pushed for the change for safety reasons.
Privacy advocates are calling for transparency about the relationship between Meta’s AI strategy and its encryption decisions. Digital Rights Watch argues that users whose private conversations may be used to train AI systems deserve to know about it. The intersection of privacy and AI is set to become an increasingly significant battleground.