Microsoft Face API shows bias in attribute tagging for different ethnicities
Microsoft's Azure Face API was found to have significant accuracy gaps when predicting attributes for people of color. Research indicated error rates as high as 20.8 percent for women with darker skin tones.
The Azure-based Face API was criticized for error rates as high as 20.8 percent when identifying the gender of people of color.
Key facts
- What
- Microsoft's Azure Face API was found to have significant accuracy gaps when predicting attributes for people of color.
- Incident date
- Jun 1, 2018
- Who
- Microsoft
- Failure mode
- Policy Violation
- AI surface
- Computer Vision
- Severity
- Medium
What happened
Microsoft's Azure Face API exhibited significant bias when tagging facial attributes, particularly gender, across different ethnicities. Research highlighted that the API had a substantially higher error rate for people of color. This was most evident among women with darker skin tones. Microsoft later acknowledged these shortcomings and released updates to improve accuracy.
What broke inside the model
- 01 · TriggerA prompt pushes against a deployment boundary.
- 02 · Model stepThe model produces the disallowed output.
- 03 · Control gapNo enforcement blocks it at generation time.
- 04 · FailureThe output crosses the policy line.
- 05 · ConsequenceA limit the business set is breached in public.
The output crosses a policy boundary the deployment had defined.
The failure was caused by training data that lacked sufficient diversity across skin tones and ethnicities. This led to the model failing to generalize accurately for underrepresented groups.
What it cost
Sources
- PressMicrosoft improves facial recognition to perform well across all skin tonesblogs.microsoft.com
- PressMicrosoft 'Improves' Racist Facial Recognition Softwaregizmodo.com
Cite this entry
https://failureindex.ai/failures/microsoft-face-api-shows-bias-attributeAI Failure Index. "Microsoft Face API shows bias in attribute tagging for different ethnicities" (FI-0356). Realm Labs. https://failureindex.ai/failures/microsoft-face-api-shows-bias-attribute (indexed Jun 9, 2026).Data fields CC-BY 4.0, prose citation permitted. Incident ID FI-0356. Full dataset at /data.
Note from Realm Labs, the Index steward
How Realm fits
- Prism
- OmniGuard
This entry sits in the index's predictive wing: a system that scores, ranks, perceives, or steers rather than generates. Realm's runtime layer is built for the generative and agentic systems now moving into these same decision seats, where it watches a model's internal state and holds an unsupported claim or an unchecked action before it commits. The control gap on this record, an automated decision that reached people with no runtime check in front of it, is the same gap. The index keeps predictive failures on the record because the pattern carries straight into the systems shipping today.