AN ACT to require the automatic analysis of police body‑worn camera recordings to flag certain behavior and improve officer training and to appropriate funds.

Whereas, less than 0.1% of police body‑worn camera data is ever reviewed despite the data being rich with insights to improve policing; and

Whereas, the sheer volume of recorded body‑worn camera footage makes it impossible for humans to analyze; and

Whereas, making body‑worn camera audio conversations searchable by utilizing natural language processing will make automated review of the body‑worn camera footage possible; and

Whereas, respectfulness and de‑escalation are interaction dimensions that can greatly enhance community trust and help repair the damaged relations between police and communities; and

Whereas, the automated review of body‑worn camera footage audio will allow for enhanced training opportunities and increased officer accountability based upon respectfulness and de‑escalation; and

Whereas, body‑worn camera analytics is the product of processing audio between law enforcement officers and civilians to generate respectfulness and de‑escalation metrics and reports; and

Whereas, these body‑worn camera analytics reports can be used to improve community policing and officer training by allowing law enforcement agencies to review the use or misuse of respectfulness and de‑escalation; Now, therefore,

The General Assembly of North Carolina enacts:

SECTION 1.  G.S. 132‑1.4A reads as rewritten:

§ 132‑1.4A.  Law enforcement agency recordings.

(a)        Definitions. – The following definitions apply in this section:

…

(1a)      Body‑worn camera analytics. – Respectfulness and de‑escalation metrics and reports generated from the processing of audio from body‑worn camera recordings between officers and civilians.

…

(4a)      Flag. – A complaint, firearm discharge, death, arrest, or investigation of an officer within a body‑worn camera recording.

(4b)      Flagging. – The process of a natural language processing technology identifying a flag during the review of a body‑worn camera recording.

(4c)      Natural language processing. – A branch of artificial intelligence that helps computers understand, interpret, and manipulate human language.

…

(b)        Public Record and Personnel Record Classification. – Recordings are not public records as defined by G.S. 132‑1. Recordings are not personnel records as defined in Part 7 of Chapter 126 of the General Statutes, G.S. 160A‑168, or G.S. 153A‑98.

(b1)      Automatically Scan and Flag Body‑Worn Camera Recordings Using Natural Language Processing Technologies. – Body‑worn camera recordings shall be searchable using artificial intelligence in order to allow automated reviews to take place and body‑worn camera analytics to be created. Natural language processing technologies may be used to review body‑worn camera recordings, identify flags, and create body‑worn camera analytics that can be used to improve community policing and officer training with a focus on the misuse of respectfulness and de‑escalation in community interactions.

Following a review of body‑camera recordings pursuant to this subsection, any video that is flagged shall be retained by the custodial law enforcement agency for no less than two years. Under no circumstances shall a flagged video be altered, hindered, obstructed, manipulated, distorted, or edited in any way.

Except as otherwise provided in this subsection, body‑worn camera analytics created pursuant to this subsection shall be retained by the custodial law enforcement agency for a minimum of 90 days. After 90 days, the body‑worn camera analytics may be retained for training purposes or destroyed.

No later than January 1, 2024, each law enforcement agency in the State that utilizes body‑worn cameras shall implement a natural language processing technology review protocol that can identify flags within body‑worn camera recordings. The natural language processing review protocol shall, at a minimum, do each of the following:

(1)        Transcribe and make searchable the audio component of all body‑worn camera recordings.

(2)        Utilize machine learning analysis or similar technology to produce daily, weekly, monthly, and yearly reports of body‑worn camera analytics for each law enforcement officer and each custodial law enforcement agency regarding the use of respectfulness and de‑escalation in all community interactions, which shall be reviewed by managers of custodial law enforcement agencies.

(3)        Send automated alerts to custodial law enforcement agency management when respectfulness and de‑escalation drops below the minimum threshold set by the North Carolina Criminal Justice Education and Training Standards Commission, indicating high levels of law enforcement officer aggression.

Each law enforcement agency in the State that utilizes body‑worn cameras shall provide, at a minimum, quarterly training regarding incidents or behavior flagged in recordings in order to address behavior that is repetitive or extreme. No later than April 1 of each year, each of these law enforcement agencies shall report to the North Carolina Criminal Justice Education and Training Standards Commission a summary of each quarterly training it provided in the previous year, any corrective action taken in the previous year to increase respectfulness and de‑escalation, and the number of law enforcement officer interactions in each of the previous two years that did not meet the minimum standards of respectfulness and de‑escalation set by the North Carolina Criminal Justice Education and Training Standards Commission pursuant to this subsection.

The North Carolina Criminal Justice Education and Training Standards Commission shall be responsible for determining the criteria of respectfulness and de‑escalation that shall be used by natural language processing technology to identify flags in body‑worn camera recordings pursuant to this subsection. In addition, the Commission shall set the threshold for when a law enforcement agency shall initiate corrective or disciplinary actions against a law enforcement officer as a result of flagged recordings found pursuant to reviews under this subsection.

….

SECTION 2.(a)  Grant Program. – There is appropriated from the General Fund to the Governor's Crime Commission within the Department of Public Safety the sum of three million dollars ($3,000,000) in nonrecurring funds for each year of the 2021‑2023 fiscal biennium to provide grants to law enforcement agencies for the purpose of purchasing natural language processing technologies and developing natural language processing technology review protocols to comply with the requirements of Section 1 of this act.

SECTION 2.(b)  Maximum Amount. – A grant provided under this section shall not exceed one hundred thousand dollars ($100,000).

SECTION 2.(c)  Guidelines. – The Governor's Crime Commission shall develop guidelines and procedures for the administration and distribution of grants under this section.

SECTION 3.  Section 1 of this act becomes effective January 1, 2024, and applies to body‑worn camera recordings made on or after that date. Section 2 of this act becomes effective July 1, 2021. The remainder of this act is effective when it becomes law.