NIST in the spotlight for facial recognition standards; some say it’s too much

When will the U.S.’ chief standards organization reach a limit in what it can do to give the facial recognition industry a coherent foundation on which to grow?

That might be unknowable, but the AI community has begun asking if the National Institute of Standards and Technology is becoming overly saturated with priorities, new lab work, governance policies and end-product reports.

Even as stated here, NIST’s responsibilities are oversimplified. Staff work on issues and standards that span biometrics. (NIST as a whole is involved in many other tech industries, too.)

An insightful analysis of the situation appears in Nextgov/FCW. Given all the facets of biometrics and other AI sectors and the high risk they pose to people, businesses and government, the article concludes the danger of operational breakdown at NIST is growing larger.

And, as if on cue, the non-profit National Academies of Sciences, Engineering and Medicine, has produced a report calling on the White House to make NIST and other agencies more proactive when it comes to all parts of facial recognition development and use.

“These are not just abstract or theoretical concerns,” according to the report, and they “have not been resolved by U.S. courts or legislatures.”

The report was sponsored by the Homeland Security Department and the Federal Bureau of Investigation. (DHS released departmental guidance on use of facial recognition in September.)

In other words, the nation needs a sharp kick in the pants to fully mitigate risks inherent to a technology that is developing without necessary “authoritative guidance, regulations, or laws,” according to another fair bit of reporting, this one from FedScoop.

The report has eight pages of detailed recommendations to prepare a foundation for effectively and ethically using facial recognition.

A top priority is having NIST “sustain a vigorous program of facial recognition technology testing and evaluation to drive continued improvements in accuracy and reduction in demographic biases.”

In fact, NIST should be “the logical home” for defining acceptable false-positive rates, evaluating key performance indicators and processing standards for quality control, data security and other areas.

The report itself is logical. Whether NIST can actually balance such a load is the question.

Article Topics

biometrics  |  demographic fairness  |  DHS  |  ethics  |  facial recognition  |  FBI  |  NIST  |  regulation