2023 - Volume 2 - Summer - Flipbook - Page 9
for some reason the trade secret information was easily
accessible (such as, for example, if a specific recipe is
given in response to a generic question about how to
make a famous soft drink) then it is possible that it
could just as easily lose its trade secret protection.
-Artificial Intelligence: Continued from page 8-
its status as a trade secret.” Id. That said, “[p]
ublication on the Internet does not necessarily destroy
the secret if the publication is sufficiently obscure or
transient or otherwise limited so that it does not become generally known” to competitors or other persons to whom the information would have economic
value. Id.; Cf. Precision Automation, Inc. v. Tech.
Svcs., Inc., No. 07-CV-707-AS, 2009 WL 116135, *2
(D. Or. April 28, 2009) (holding that posting of information on company’s website, even if briefly, rendered it not a trade secret).
Takeaway
The key takeaway for businesses keen on balancing the risk and reward of generative AI technology is
to understand that the “secrecy” of information will be
determined by the context of any disclosure and that
mere posting of the information online may not be
enough to destroy any protection. Still, if a generative
AI model is trained on the information and somehow
manages to easily spit it out in response to benign requests, it may lose its status as a trade secret.
Another instructive context is in cases dealing
with the publication of trade secret information on
things like court documents. See, e.g., Kittrich Corp.
v. Chilewich Sultan, LLC, No.
CV1210079GHKARGX, 2013 WL 12131376 (C.D.
Cal. Feb. 20, 2013); Hurry Fam. Revocable Tr. v.
Frankel, No. 8:18-CV-2869-CEH-CPT, 2023 WL
23805 (M.D. Fla. Jan. 3, 2023; The Equal Rights Center v. Lion Gables Residential Trust, No. DKC 072358, 2010 WL 2483613, *3 (D. Md. June 15, 2010);
HMS Holdings Corp. v. Arendt, 18 N.Y.S.3d 579
(Table), *8 (N.Y. Sup. Ct. Albany Cnty. 2015).
Businesses should always read and understand the
scope of any end-user license agreement provided by
the vendor of any generative AI applications it or its
employees may use. Many of these agreements used
by AI companies allow the company to review, release, or even sell sensitive information shared with it.
These generative AI applications also almost uniformly contain unilateral confidentiality provisions, binding
the user but allowing the AI purveyor free reign
(besides privacy law constraints) to use information
shared with it.
For example, in Frankel, the court explained that,
though information that was arguably a trade secret
was posted on the court’s electronically available
docket, because the publication was obscure or otherwise limited, it did not destroy trade secret protection.
Even though the information was technically publicly
available, it was not easy to access, and Plaintiff’s
competitors would struggle to locate it. Members of
the public would only be able to find the information
if they knew specific information about the case because the relevant information was unlabeled and located within a docket entry that contained numerous
attachments. These impediments to easy access outweighed the fact that the information was publicly
available through a searching inquiry.
These risks should remain top of mind for businesses intent on capitalizing on the benefits of generative AI.
EMPLOYMENT LAW
Increasing Federal Attention
On April 25, 2023, four federal agencies released a
“joint statement on enforcement efforts against discrimination and bias in automated systems.” Joint
Statement on Enforcement Efforts Against Discrimination and Bias in Automated Systems. The joint statement expressed commitments by the Consumer Financial Protection Bureau, the Department of Justice’s
Civil Rights Division, the Equal Employment Opportunity Commission, and the Federal Trade Commission to “ensure that these rapidly evolving automated
systems are developed and used in a manner consistent
with federal laws.” Id. at 2. The joint statement also
provided links to guidance documents prepared by
each agency explaining their enforcement roles in dif-
Applying the reasoning of Bunner and Frankel
above to the context of generative AI, we can draw
some conclusions about how a court might rule on the
issue. Information inputted into a generative AI model
has the potential to be made publicly available. Even
so, the information cannot be easily viewed or disseminated by third parties. Because of the various layers required to potentially view this information, it is
possible that a court could find any trade secrets disclosed to a generative AI model are—like information
buried on a court docket—still protected. That said, if
-Continued on page 10-
9
