Full text
- This is an unofficial copy. The document has been archived and reformatted in plaintext for AGORA. Footnotes, tables, and similar material may be omitted. For the official text, visit the original source.
- Thematic tags for this document are awaiting validation (peer review by a second AGORA editor).
SEC. 1513. PHYSICAL AND CYBERSECURITY PROCUREMENT REQUIREMENTS FOR
ARTIFICIAL INTELLIGENCE SYSTEMS.
(a) Security Framework.--
(1) In general.--The Secretary of Defense shall develop a
framework for the implementation of cybersecurity and physical
security standards and best practices relating to covered
artificial intelligence and machine learning technologies to
mitigate risks to the Department of Defense from the use of such
technologies.
(2) Coverage of relevant aspects of security.--The framework
developed under paragraph (1) shall cover all relevant aspects of
the security of artificial intelligence and machine learning
systems of the Department of Defense, including the following:
(A) Risk posed to and by the workforce of the Department of
Defense, including insider threat risks.
(B) Training and workforce development requirements,
including with respect to the following:
(i) Artificial intelligence security awareness.
(ii) Artificial intelligence-specific threats and
vulnerabilities.
(iii) Development of a continuum of professional
development and education of artificial intelligence
security expertise.
(C) Risks to the supply chains of such systems, including
counterfeit parts or data poisoning risks.
(D) Risks relating to adversarial tampering with artificial
intelligence systems.
(E) Risks relating to the unintended exposure or theft of
artificial intelligence systems or data.
(F) Security posture management practices, including
governance of security measures, continuous monitoring, and
incident reporting procedures.
(G) An evaluation of commercially available platforms for
continuous monitoring and assessment of such systems.
(3) Risk-based framework.--The framework developed under
paragraph (1) shall be risk-based, including security that is
proportional to the national security or foreign policy risks posed
by the covered artificial intelligence and machine learning
technology being stolen or tampered with.
(4) Use of existing frameworks.--To the maximum extent
feasible, the framework developed under paragraph (1) shall--
(A) draw on existing cybersecurity reference documents,
including the NIST Special Publication 800 series; and
(B) be implemented as an extension or augmentation of
existing cybersecurity frameworks developed by the Department
of Defense, including the Cybersecurity Maturity Model
Certification framework.
(5) Addressing extreme security risks.--
(A) Highly capable cyber threat actors.--The framework
developed under paragraph (1) shall prioritize the most highly
capable artificial intelligence systems that may be of highest
interest to cyber threat actors, based on risk assessments and
threat reporting.
(B) Security levels.--The Secretary shall ensure that the
framework developed under paragraph (1) imposes requirements
for security on contractors that are designed to mitigate the
cyberesecurity risks posed by the cyber threat actors described
in subparagraph (A), with the most stringent security
requirements under such frameworks providing protection that is
similar to the protection offered by national security systems
(as defined in section 3552(b)(6) of title 44, United States
Code).
(C) General design with specific components.--To the extent
feasible, any additional security requirements developed
pursuant to subparagraph (B) shall be designed generally for
all software systems of the Department of Defense, but may
contain components designed specifically for highly capable
artificial intelligence systems.
(b) Security Requirements.--
(1) In general.--The Secretary of Defense shall amend the
Defense Federal Acquisition Regulation Supplement, or take other
similar action, to require covered entities to implement the best
practices described in subsection (a) under the framework developed
under such subsection.
(2) Risk-based rules.--Any requirements implemented pursuant to
paragraph (1) shall, to the extent practicable, be narrowly
tailored to the specific covered artificial intelligence and
machine learning technologies developed, deployed, stored, or
hosted by a covered entity, and shall be calibrated accordingly to
the different tasks involved in development, deployment, storage,
or hosting of components of such covered artificial intelligence
and machine learning technologies.
(3) Cost-benefit consideration.--
(A) In general.--In carrying out paragraph (1), the
Secretary of Defense shall--
(i) consider the costs and benefits to the Department
of Defense and to the national security and technological
leadership of the United States, of imposing security
requirements on covered entities; and
(ii) to the extent feasible, design the requirements
implemented pursuant to such paragraph to allow for trade
space analysis by the Department in a transparent manner
between competing requirements in order to minimize the
costs and maximize the benefits of such requirements.
(B) Weighing costs of slowing down development.--In
carrying out subparagraph (A), the Secretary shall weigh the
costs of slowing the development and deployment of artificial
intelligence and machine learning against the benefits of
mitigating national security risks and potential security risks
to the Department of Defense from using commercial software for
imposing additional physical or cybersecurity requirements for
such systems.
(c) Private Sector Collaboration.--In carrying out the requirements
of subsection (a), the Secretary of Defense shall seek to collaborate
with industry and academia in the development of the framework under
such subsection using a process for consultation that uses a new or
existing mechanism for public-private partnerships.
(d) Implementation Plan.--The framework required by subsection
(a)(1) shall include a detailed plan for the implementation of the
framework that--
(1) establishes timelines and milestones for achieving the
objectives outlined in the framework;
(2) identifies resource requirements and funding mechanisms;
and
(3) provides metrics for measuring progress and effectiveness.
(e) Reporting Requirements.--Not later than 180 days after the date
of the enactment of this Act, the Secretary shall submit to the
congressional defense committees an update on the status of
implementation of the requirements of this section.
(f) Definitions.--In this section:
(1) The term ``artificial intelligence'' has the meaning given
such term in 238(g) of the John S. McCain National Defense
Authorization Act for Fiscal Year 2019 (Public Law 115-232; 10
U.S.C. 4061 note prec.).
(2) The term ``covered artificial intelligence and machine
learning technology'' means an artificial intelligence or machine
learning system acquired by the Department of Defense or an element
of the Department and all associated components involved in the
development and deployment lifecycle of such system, including
source code, numerical parameters (including model weights) of the
trained artificial intelligence or machine learning system, details
of any methods and algorithms used to develop such system, data
used in the development of such system, and software used for
evaluating the trustworthiness of the artificial intelligence or
machine learning system during development or deployment.
(3) The term ``covered entity'' means an entity that enters
into a contract or other agreement with the Department of Defense
under which such entity engages in the development, deployment,
storage, or hosting of one or more covered artificial intelligence
and machine learning technologies.