Elaine Moore Published October 1 2026
AI doomers may have privacy worriers in a headlock, but there’s one identifier that remains stubbornly beyond artificial reach: smell. As a data modality, olfactory signals have seen little AI progress, leaving a niche but meaningful gap in surveillance capabilities. OpenAI contractors reportedly read ChatGPT conversations, and Meta’s AI agent Muse is said to bypass permissions, making anonymity feel increasingly obsolete.
Barry Smith, director of the Institute of Philosophy at the University of London’s School of Advanced Study, argues that smell preserves a boundary between real and virtual worlds. “I have attributes not digitally coded,” he says. Without scent, even immersive virtual environments or digitally recorded memories fall short of true realism. Smith emphasizes that taste and smell anchor us to lived experience, resisting full digital replication.
Yet researchers like Alex Wiltschko, founder of Osmo and former Google Brain scientist, believe AI noses are inevitable. Comparing olfactory intelligence to early photography, he notes the challenge lies in quantifying ephemeral scents. Building data labeling infrastructure — akin to a “Rosetta Stone” for molecules — is key. MIT’s Smell Net and efforts at Rockefeller University highlight the scale of the task. For now, however, the human nose remains an unconquered frontier for AI.
Source: Financial Times Companies · Summarized by HeadlinesBriefing