Algorithm Accountability Act
- Sponsor
- Sen. Curtis, John R. [R-UT]
- Committees
- Commerce, Science, and Transportation Committee (primary)
- Last action
- Nov 18, 2025
Bottom line
This bill seeks to hold large social media companies accountable for physical harm caused by their algorithmic recommendations by limiting their Section 230 immunity and creating a pathway for civil lawsuits.
What it actually does
This bill amends Section 230 of the Communications Act of 1934 to establish a new duty of care for large social media platforms regarding their recommendation-based algorithms. It makes platforms liable for bodily injury or death foreseeably caused by these algorithms, removing Section 230 immunity in such cases and establishing a private right of action, while also invalidating predispute arbitration agreements for these disputes.
Proponents argue
Proponents argue that current law, specifically Section 230, provides overly broad immunity that shields social media companies from accountability for the real-world harms their algorithms can amplify or cause, such as promoting self-harm content or inciting violence. They contend that platforms should bear responsibility for the design choices of their powerful recommendation systems, especially when those choices lead to foreseeable physical harm or death.
Opponents contend
Opponents might argue that this bill could stifle innovation in algorithmic design, lead to overly cautious content moderation that suppresses legitimate speech, and potentially open platforms to a flood of litigation that is difficult to defend against. They may also raise concerns that determining 'reasonable care' for algorithms is inherently complex and could lead to inconsistent judicial interpretations, ultimately harming the open nature of the internet.
The bill is relatively short, but its implications for existing legal frameworks and the tech industry are substantial, requiring careful consideration beyond a quick read.
Section 2(a)(2), new subsection (f)(1)(A)
Establishes Algorithmic Duty of Care
This provision requires large social media platforms to act with 'reasonable care' when designing, testing, deploying, and maintaining their recommendation algorithms. The goal is to prevent foreseeable bodily injury or death that can be linked to how these algorithms operate. This means platforms must proactively consider and mitigate risks associated with their algorithmic content promotion.
Supporters argue
Supporters argue that this duty is essential to modernize liability law for the digital age, ensuring that powerful algorithms are designed with public safety in mind, much like other products. They believe it will incentivize platforms to prioritize user well-being over engagement metrics that can sometimes lead to harmful content amplification.
Critics contend
Critics may contend that defining 'reasonable care' for algorithms is ambiguous and could lead to subjective interpretations, creating a chilling effect on innovation. They might also argue that it places an undue burden on platforms to predict and prevent all potential harms, which are often complex and multifactorial.
Tradeoffs
The bill balances the desire to hold platforms accountable for algorithmic harms against concerns about stifling innovation, potential over-moderation of content, and the practical challenges of proving causation in complex digital environments.
Section 2(a)(2), new subsection (f)(2)
Establishes Enforcement Mechanisms
If a social media platform violates its duty of care, it loses its Section 230(c)(1) immunity, meaning it can be held liable for content recommendations. Furthermore, individuals who suffer bodily injury or death due to such a violation can sue the platform in federal court for compensatory and punitive damages. This creates a direct legal pathway for victims to seek redress.
Section 2(a)(2), new subsection (f)(3)
Prohibits Forced Arbitration for Algorithmic Harms
This section makes any predispute arbitration agreements or joint-action waivers invalid and unenforceable for disputes arising under the new algorithmic accountability provisions. This means users cannot be forced into arbitration and retain their right to pursue claims in court, including class action lawsuits, for harms related to algorithmic design.
Section 2(a)(2), new subsection (f)(6)
Defines Covered Entities and Algorithms
The bill carefully defines which entities are considered 'social media platforms' subject to the new duty of care, generally focusing on large, for-profit interactive computer services where users primarily interact with content. It explicitly excludes smaller platforms (under 1 million users), direct messaging services, internal corporate communication tools, teleconferencing, and platforms primarily for reviews, e-commerce, streaming, or news. It also defines 'recommendation-based algorithm' as automated systems that curate content based on user personal data.
This provision explicitly states that the new duty of care does not apply to content sorted chronologically or reverse chronologically, nor to the initial results of a user-initiated search. However, it clarifies that the exception for search results does not apply once a user navigates beyond those initial results.
Section 2(a)(2), new subsection (f)(1)(C)
Why it matters:This is standard legislative practice for defining the precise scope of a new rule, ensuring it doesn't inadvertently cover benign or non-problematic functionalities. It helps to clarify the bill's intent.
Case for: Supporters would argue this exception is vital to ensure the bill targets recommendation algorithms specifically, which are designed to proactively push content, rather than passive display methods like chronological feeds or direct search queries where user intent is clear. It prevents overreach and protects basic internet functionalities.
Case against: Critics might argue that even chronological feeds or initial search results can be manipulated or designed in ways that lead to harm, and that carving out these exceptions could create loopholes for platforms to exploit.
Estimated impact: This significantly narrows the scope of the bill, focusing liability squarely on proactive, personalized recommendation systems rather than all forms of content display.