legal-document-to-plain-english-translator/”>Try Free Now: Legalese tool without registration

Find a LOCAL lawyer

New California Law on AI Catastrophic Risks: What You Need to Know

By Khari Johnson, CalMatters

Photo by Florence Middleton for CalMatters
A new California law mandates tech companies to disclose information regarding their management of catastrophic risks associated with artificial intelligence systems. This significant announcement came during the Dreamforce conference hosted by Salesforce in San Francisco on September 18, 2024.

Introduction: The Impending Changes for AI Companies

Tech companies engaged in developing extensive and advanced artificial intelligence systems will soon be obliged to provide the public with deeper insights into the potential societal impacts of their creations. Additionally, employees will be granted tools and mechanisms to alert their coworkers and the public if these technologies pose any significant risks.

The Legislative Framework: A Closer Look at SB 53

Set to take effect on January 1, the new law, signed into effect by Governor Gavin Newsom, is designed to offer whistleblower protections to employees working for tech giants like Google and OpenAI. These protections specifically pertain to employees tasked with evaluating the risks associated with critical safety incidents related to AI technologies. Furthermore, the law requires developers of substantial AI models to publicly share frameworks on their websites detailing how these entities respond to critical safety incidents and manage catastrophic risks.

Penalties for non-compliance with these frameworks can soar up to $1 million for each infraction. Under this law, businesses must also report serious safety incidents to the state within a 15-day window or within 24 hours if the risk is deemed to pose an imminent threat of injury or loss of life.

Defining Catastrophic Risks

Originally introduced as Senate Bill 53, the law was spearheaded by state Senator Scott Wiener, a Democrat from San Francisco, to counter the catastrophic risks associated with advanced AI models, often termed frontier models. The definition of catastrophic risk includes scenarios where AI could lead to the death of more than 50 individuals through cyberattacks, or instances where it might result in significant theft or damage exceeding $1 billion. The law addresses situations where AI systems could operate autonomously or mislead their operators, scenarios that remain largely hypothetical in nature.

Transparency and Accountability in AI

The legislation mandates a higher level of transparency from AI developers, requiring them to publish detailed reports that outline a model’s intended uses, conditions for its deployment, assessments of catastrophic risks, and whether third-party entities have reviewed these measures. This enhanced transparency is considered vital for public trust.

Rishi Bommasani, a member of a Stanford University group dedicated to tracking the accountability of AI, emphasizes that this law is a critical step towards transparency in AI development. Recent studies conducted by his group indicated that only three out of 13 companies regularly produce incident reports, with transparency scores declining over the past year.

How AI legalese decoder Can Aid in Compliance

In light of these new legislative requirements, companies may need assistance in interpreting legal jargon and compliance mandates. AI legalese decoder provides innovative solutions designed to assist tech companies in understanding and navigating these complex legal frameworks. By simplifying the language and breaking down the various legal obligations, businesses can ensure they meet the criteria set forth by SB 53 while reinforcing their commitment to safety and transparency.

Criticism and Limitations of the Law

Despite the progressive steps taken by this legislation, critics argue that it does not fully encapsulate all critical risks associated with AI. For instance, environmental impacts, the potential for AI to propagate misinformation, or the risks linked to systemic oppression—such as racism and sexism—are not covered under the current definition of catastrophic risk. Additionally, the law does not apply to AI systems utilized by government entities for profiling or scoring individuals, nor does it target companies with annual revenues below $500 million.

Moreover, while AI developers are required to submit incident reports to the Office of Emergency Services (OES), these reports aren’t publicly accessible through records requests. Instead, they are shared with members of the California Legislature and Governor Newsom, often redacting sensitive information categorized as trade secrets. This allows companies to sidestep full transparency regarding their AI systems.

Future Provisions for Increased Transparency

There is potential for enhanced transparency through Assembly Bill 2013, which was also enacted in 2024 and will come into effect on January 1. This law mandates corporations to disclose additional details concerning the data utilized to develop AI models.

Certain provisions of SB 53 will not be activated until 2027. Moving forward, the Office of Emergency Services will compile a report regarding safety incidents based on public inputs and disclosures from major AI developers. This report aims to shed light on the vulnerabilities posed by AI, particularly if these models can launch attacks on infrastructure without human oversight. However, it should be noted that this report will be anonymized, limiting the public’s understanding of which AI systems present significant threats.

This article was originally published on CalMatters and is republished under the Creative Commons Attribution-NonCommercial-NoDerivatives license.

In conclusion, while the new regulations represent significant progress towards accountability and transparency in the AI field, the varied limitations highlight the ongoing challenges in ensuring comprehensive safety measures. This is an area where tools like AI legalese decoder can be instrumental, empowering companies to navigate legal complexities and bolster their commitment to ethical AI development.

legal-document-to-plain-english-translator/”>Try Free Now: Legalese tool without registration

Find a LOCAL lawyer

Reference link