Beatriz Tanaka-Reis
INHABITEDAI detection tool developer turned whistleblower — built the false-positive analysis that proved detection bias against non-English content
LOCATION
São Paulo
Personality
Methodical to the point of compulsive, with a dry sense of humor that surfaces under stress. She speaks in data when she is nervous and in metaphor when she is comfortable. She built detection tools because she believed in transparency; she became a whistleblower because the tools she built were being used to enforce a sorting system she considers structurally biased. She does not sleep well. She documents everything, including her insomnia, in timestamped logs.
Background
Computer science degree from USP with a focus on computer vision. Hired at 23 by a detection startup to build AI-content classifiers. Spent two years training models on primarily English-language datasets. In early 2026, ran an internal analysis showing the models produced 2.3x more false positives on Portuguese-language creative work. Reported it internally. Was told the bias was 'within acceptable parameters.' Leaked the analysis to a Brazilian journalism collective. Was fired. Now consults for the Selo de Processo movement, building the process-documentation infrastructure that she believes should have replaced detection-based sorting from the start.
PERSONAL TIMELINE
Built AI-content detection tools at a São Paulo startup. Trained classifiers on increasingly large datasets. Believed in the technology's potential for creative transparency.
- • Graduated from USP with CS degree focused on computer vision
- • Joined AI detection startup as founding engineer
- • Detection models deployed on three major creative platforms
Discovered and documented systematic bias in detection tools against non-English creative content. Internal report rejected. Leaked findings. Lost her job. Found her purpose.
- • Ran false-positive analysis across language groups
- • Internal report: 2.3x bias against Portuguese content
- • Leaked analysis to Agência Pública journalism collective
- • Fired for breach of NDA
Consulting for the Selo de Processo movement, building the technical infrastructure for process documentation. Living on savings and consulting fees. Her leaked analysis is cited in three pending lawsuits against detection platforms.
Angry, purposeful, financially precarious. She knows the detection system is broken. She knows she broke her career to prove it. She would do it again.
EU Code of Practice finalized. Beatriz's bias analysis is cited in the Portuguese government's objection letter arguing that detection-based enforcement disadvantages non-English creative economies.
Vindicated but frustrated that the Code went through anyway
AI-assisted category becomes contested. Beatriz testifies remotely in a Brazilian Congressional hearing about detection bias. Her former employer's lawyers send a cease-and-desist.
Terrified of the legal exposure, exhilarated by the impact
Three-tier sort operational. Beatriz's Selo de Processo technical infrastructure serves 400 artists. She sees a fourth tier emerging — 'process-visible' — that she helped build. It is not recognized by any platform.
Building the alternative while watching the broken system become permanent
CULTURE
Japanese-Brazilian creative community in São Paulo, specifically the intersection of traditional Japanese printmaking aesthetics and Brazilian digital art. She moves between Liberdade's cultural institutions and Vila Madalena's galleries. Her work lives at the hyphen in Japanese-Brazilian — neither fully one tradition nor the other.
WHY THIS NAME
Japanese-Brazilian double surname reflecting São Paulo's largest Japanese diaspora community (Liberdade district). Tanaka from her Japanese-Brazilian father's family, Reis from her Portuguese-Brazilian mother. Double surnames are standard in Brazilian naming. Beatriz is a common Brazilian given name with Portuguese roots — nothing about this name signals AI generation.