Trains models, in the United States

17 U.S.C. § 102 · Read the law's page

• Treat an original work of authorship fixed in a tangible medium of expression, in one of the listed categories such as literary works, as protected by copyright (s. 102(a)).
• Do not assume that copyright in a computer program protects an idea, procedure, process, system, method of operation, concept, principle or discovery it embodies, in whatever form it is described, explained, illustrated or embodied (s. 102(b)).
• Read section 102(b) for a computer program as the House report does: the expression the programmer adopted is the copyrightable element, and the actual processes or methods the program embodies are outside the scope of the copyright law.

17 U.S.C. §§ 105, 302 and 511 · Read the law's page

• Do not assume you need the Government's permission to reproduce a work of the United States Government: copyright protection under title 17 is not available for it (s. 105(a)).
• Get permission from the owner before you reproduce a copyright that the United States Government holds by transfer, whether by assignment, bequest, or otherwise (s. 105(a)).
• Treat the covered author, not the Government, as the owner of the copyright in a literary work produced by a civilian faculty member of a covered institution in the course of employment there for publication by a scholarly press or journal (s. 105(b), (d)).
• Count a work created on or after January 1, 1978 as protected for the life of the author and 70 years after the author's death, and a joint work not made for hire for the life of the last surviving author and 70 years after that author's death (s. 302(a), (b)).
• Count an anonymous work, a pseudonymous work, or a work made for hire as protected for 95 years from the year of its first publication or 120 years from the year of its creation, whichever expires first (s. 302(c)).
• Do not rely on section 511(a) to sue a State or State instrumentality for infringement: the Supreme Court held that Congress lacked authority to abrogate the States' immunity from copyright infringement suits in the Copyright Remedy Clarification Act (Allen v. Cooper (2020)).

17 U.S.C. §§ 106 and 501 · Read the law's page

• Get the copyright owner's permission before you reproduce a copyrighted work in copies or phonorecords, prepare a derivative work based on it, or distribute copies or phonorecords of it to the public by sale or other transfer of ownership, or by rental, lease, or lending, unless sections 107 through 122 allow the use (s. 106(1) to (3)).
• Get the copyright owner's permission before you publicly perform or publicly display a work of a kind clauses (4) and (5) of section 106 name, or publicly perform a sound recording by means of a digital audio transmission, unless sections 107 through 122 allow the use (s. 106(4) to (6)).
• Treat transmitting or otherwise communicating a performance or display of a work to the public by any device or process as performing or displaying it publicly, whether the recipients receive it in the same place or in separate places and at the same time or at different times (s. 101).
• Do not violate any of the exclusive rights provided by sections 106 through 122, or import copies or phonorecords into the United States in violation of section 602: anyone who does is an infringer, including a State, a State instrumentality, and a State officer or employee acting in an official capacity (s. 501(a)).
• Do not induce or encourage users to infringe, and do not profit from their infringement while declining to exercise a right to stop or limit it: the Supreme Court describes the first as contributory infringement and the second as vicarious infringement (MGM Studios v. Grokster, 545 U.S. 913 (2005)).

17 U.S.C. §§ 109 and 117 · Read the law's page

• You may sell or otherwise dispose of the possession of a particular copy or phonorecord you own that was lawfully made under title 17 without the copyright owner's authority; this does not extend, unless the copyright owner authorizes it, to a person who acquired possession from the copyright owner by rental, lease, loan or otherwise without acquiring ownership of it (s. 109(a) and (d)).
• Do not dispose of the possession of a copy of a computer program by rental, lease or lending, or by any act in the nature of rental, lease or lending, for direct or indirect commercial advantage unless the owner of copyright in the program authorizes it (s. 109(b)(1)(A)).
• That rental limit does not reach a program embodied in a machine or product that cannot be copied during its ordinary operation or use, a program in a limited purpose computer designed for playing video games, or lending for nonprofit purposes by a nonprofit library that affixes the prescribed copyright warning to each copy (s. 109(b)(1)(B) and (2)(A)).
• You may make another copy or adaptation of a computer program you own a copy of, without infringing, only as an essential step in using the program with a machine and for no other use, or for archival purposes only, and you must destroy all archival copies if continued possession of the program ceases to be rightful (s. 117(a)).
• Transfer exact copies made under section 117 only along with the copy they were made from and only as part of the transfer of all rights in the program, and transfer adaptations made under it only with the copyright owner's authorization (s. 117(b)).
• The owner or lessee of a machine may make a copy of a program solely by virtue of activating the machine that lawfully contains an authorized copy, for maintenance or repair of that machine only, if the copy is used in no other manner, is destroyed immediately after the maintenance or repair is completed, and any part of the program not necessary to activate the machine is not accessed or used other than to make the copy (s. 117(c)).

17 U.S.C. § 1202 · Read the law's page

• Do not intentionally remove or alter copyright management information, meaning the title, the names of the author and copyright owner, the copyright notice and the terms and conditions for use, on works you copy, unless the copyright owner or the law authorizes it.
• Do not distribute, import for distribution or publicly perform works or copies, or distribute the information itself, knowing that copyright management information has been removed or altered without authority; for works and copies the text also requires that you know, or for civil remedies have reasonable grounds to know, that doing so will induce, enable, facilitate or conceal an infringement.
• Do not provide or distribute false copyright management information with the intent to induce, enable, facilitate or conceal infringement.

412 F. Supp. 2d 1106 (D. Nev. 2006)

• Expect caching of web pages for search results to be defensible: the District of Nevada granted Google summary judgment, holding that it did not directly infringe, that the caching was fair use, and that the DMCA's system-cache safe harbor barred damages.
• Expect an implied license to be found where a site owner knew it could stop cached links with the 'no-archive' meta-tag and did not use it: the court so held against this author and held him estopped from suing.
• Do not assume the same result where a site has used the 'no-archive' meta-tag.

31 F.4th 1180 (9th Cir. 2022)

• Expect the Computer Fraud and Abuse Act case against scraping public webpages to be contested: affirming a preliminary injunction for the scraper, the Ninth Circuit found a serious question whether 'without authorization' reaches only computers where authorization, such as password authentication, is generally required.
• Do not read the ruling as clearing scraping under other laws: the court noted that state-law trespass to chattels claims may still be available, and that copyright infringement, misappropriation or breach of contract claims may also lie.

28 CFR Part 202 · Read the law's page

• Do not engage in a prohibited bulk data-brokerage transaction involving bulk U.S. sensitive personal data, including biometric identifiers such as facial images or voice prints, or government-related data, with a country of concern or covered person.
• Apply the required security safeguards to a restricted vendor, employment, or investment transaction that would give a country of concern or covered person access to that data.
• Treat data as in scope even when it is anonymized, pseudonymized, de-identified or encrypted: a collection of sensitive personal data counts toward the bulk threshold in any of those forms (28 CFR 202.206).

Discussion Draft, 119th Cong. (June 4, 2026)

H.R. 3460, 119th Congress · Read the law's page

• Do not discharge, demote, suspend, threaten, blacklist, harass, or otherwise discriminate against an employee or independent contractor, current or former, in the terms and conditions of employment or work, because of a lawful act in reporting an AI security vulnerability or AI violation, or conduct the person reasonably believes is one (Section 3(a)).
• Treat as protected a report to the appropriate regulatory official, the Attorney General, a regulatory or law enforcement agency, a Member or committee of Congress, a person with supervisory authority at the employer, or another individual at the employer whom the person reasonably believes can investigate or stop the misconduct, and the person's part in an investigation or action based on it (Section 3(a)(1) to (3)).
• Treat an AI violation as any violation of Federal law related to or committed during the development, deployment, or use of artificial intelligence, or a failure to respond appropriately to a substantial and specific danger that artificial intelligence may pose to public safety, public health, or national security (Section 2(2)).
• Treat an AI security vulnerability as any failure or lapse in security that could potentially allow emerging artificial intelligence technology to be acquired by a person, including a foreign entity, by theft or other means (Section 2(1)).
• Do not use a contract, agreement, policy form, or condition of employment, including a requirement to arbitrate or mediate first, to waive or alter the whistleblower's rights and remedies (Section 3(c)).

H.R. 9915, 119th Congress · Read the law's page

• Do not deploy a bot that accesses a website, digital platform, or online service without disclosing its identity and purpose, in a manner reasonably likely to damage, impair, or burden the technical or commercial operation of the website or service.
• Identify your bot through a valid and accurate user-agent string, and disclose its specific nature and purpose, including any proposed use of page content such as search indexing or artificial intelligence training, at the time access is requested and in a format the website operator can access.
• Do not intentionally misrepresent, conceal, or obscure the nature of a bot so that it appears to be a human user, for use in connection with a generative AI model or service.

H.R.7209 / S.2455, 119th Cong. (2025-2026)

S. 2367, 119th Congress · Read the law's page

• Do not appropriate, use, collect, process, sell, or otherwise exploit an individual's covered data, in or affecting interstate or foreign commerce, without the individual's express, prior consent (Section 3(a)).
• Do not train a generative artificial intelligence system that you sell, rent, license, or otherwise use on an individual's covered data without that individual's express, prior consent, and do not generate covered data pertaining to an individual, including content that imitates, replicates, or is substantially derived from the individual's covered data, without it (Section 2(1) with Section 3(a)).
• Treat covered data as information that can reasonably be linked to a specific individual, information derived, inferred, or generated from it, and material the individual created that is protected by copyright whether or not it is registered (Section 2(4)).
• Obtain consent as a clear, affirmative act made in advance, freely given, informed, and unambiguous, for the specific use (Section 2(5)); consent obtained through coercion or deception, or as a condition of using a product or service where the use exceeds what is reasonably necessary to provide it, is not valid (Section 3(b)(3)(B)).
• Where covered data goes to or is used by a third party, specifically and clearly disclose each third party to the individual when consent is sought, present the disclosure distinctly and separately from any privacy policy or terms of service, and do not rely on a hyperlink or a general reference to a privacy policy (Section 3(d)).
• Do not rely on a predispute arbitration agreement or a predispute joint-action waiver to avoid a claim under the Act; neither is valid or enforceable for it (Section 3(c)).