BILL NUMBER: S8331A
SPONSOR: GONZALEZ
 
TITLE OF BILL:
An act to amend the general business law, in relation to requiring tran-
sparency from generative artificial intelligence developers for journal-
ism providers
 
PURPOSE:
This bill is to provide transparency in the utilization of publication
information used by generative AI developers.
 
SUMMARY OF PROVISIONS:
Section one: New York artificial intelligence transparency for journal-
ism act.
Section two: Legislative findings.
Section three: Establishes a process by which developers of artificial
intelligence systems provide for disclosure when they utilize print,
broadcast, broadcast network or digital publication or service journal-
istic work product to train generative AI systems.
Section four: Effective date.
 
JUSTIFICATION:
The purpose of this legislation is to provide public disclosure when
journalistic work product is used to train generative AI systems.
Broadcasters and news publishers invest millions of dollars annually to
gather and report local news in communities throughout the New York.
Technology companies gather this original news content, usually without
consent or compensation, and use it to train their large language AI
systems in order to generate summaries of the gathered information. AI
developers use technological tools to scrape original reporting from
behind news organizations' paywalls and to evade technologies designed
to prevent unauthorized copying of news content.
This legislation is narrowly crafted to allow news publishers and broad-
casters ability to know when their content is being used to train and
fine tune large language AI systems. It simply requires transparency by
AI developers. Pursuant to the legislation, technology companies would
be required to maintain a publicly available website listing the news
sources they have used to train their AI systems.
The legislation is fully consistent with federal copyright law. The
focus of the legislation is simply to provide transparency. Liability is
imposed on an AI developer solely for failing to list the news sources
it has used to train and fine tune its large language AI model. This is
distinct from copyright liability, which focuses on the use of such
content. This legislation does not address whether using content to
train large language AI systems constitutes "fair use" under federal
copyright law-it only requires disclosure.
 
LEGISLATIVE HISTORY:
This is a new bill.
 
FISCAL IMPLICATIONS:
None
 
EFFECTIVE DATE:
This Act shall take effect immediately.