2023-06-24-美国电子隐私信息中心-2023生成式人工智能的潜在有害影响与未来之路（英文版）_86页_13mb

Summary of "Generating HARMs: Generative AI's Impact & Paths Forward"

Core Content

This document, Generating HARMs: Generative AI's Impact & Paths Forward, published in May 2023, outlines the significant risks and harms associated with the rapid deployment of generative AI technologies. It emphasizes the need for regulatory and legal interventions to address these harms, which include misinformation, disinformation, cybersecurity threats, harassment, and discrimination. The report is authored by a group of experts and edited by Grant Fergusson, Calli Schroeder, Ben Winters, and Enid Zhou, with contributions from several scholars.

Main Viewpoints

The report presents the following key viewpoints:

• Generative AI is a double-edged sword: While it offers innovative capabilities, it also enables the creation and spread of harmful content at an unprecedented scale and speed.
• Privacy and consent are undermined: AI-generated content, especially deepfakes, often uses publicly available information, making traditional privacy torts less applicable and complicating legal redress.
• Legal frameworks are insufficient: Current laws, such as those related to defamation, cyberstalking, and privacy, are not well-equipped to handle the unique challenges posed by generative AI.
• Self-regulation has failed: The lack of effective industry self-regulation has led to a proliferation of harmful AI applications, necessitating stronger legislative and regulatory responses.
• Public trust is at risk: As AI-generated content becomes more realistic, the public may increasingly be misled, leading to broader societal and democratic concerns.

Key Information

Types of Harm

The report categorizes AI-generated harms into several types, including:

• Economic/Economic Loss: Scams and malware can lead to direct financial loss and long-term credit damage.
• Reputational/Relationship/Social Stigmatization: False information can damage an individual's reputation and relationships.
• Psychological/Emotional Distress: Victims of disinformation or deepfakes may experience shame, embarrassment, or emotional harm.
• Autonomy: Misinformation and surveillance advertising can manipulate user behavior and decision-making.
• Discrimination: AI tools can exploit vulnerabilities of certain groups, such as the elderly, immigrants, or those with disabilities.

Harmful Applications

• Scams: AI can be used to create convincing phishing emails and robocalls, targeting vulnerable populations.
• Disinformation: AI facilitates the spread of false or misleading content, which can influence public opinion and elections.
• Misinformation: AI-generated text may contain inaccuracies that are difficult to detect, especially when presented alongside factual information.
• Cybersecurity Threats: AI can assist in the creation and modification of malware, increasing the risk of cyberattacks.
• Clickbait and Surveillance Advertising: AI-generated content can be used to manipulate user engagement and exploit personal data for targeted advertising.

Case Studies

• Election 2024: Generative AI can create convincing scripts and videos to spread election misinformation, posing a threat to democratic processes.
• Silencing a Journalist: Deepfakes were used to falsely portray a journalist in a sexual context, leading to public humiliation and threats.

Legal Challenges

• Malicious Intent: Many criminal statutes require proof of intent, which is difficult to establish with AI-generated content.
• Privacy and Consent: AI-generated images and videos often use public material, making traditional privacy claims less effective.
• First Amendment Concerns: The line between free expression and privacy violations is blurred, especially when AI-generated content appears real.

Recommendations

The report suggests the following interventions:

• Enact legislation to prohibit the use of AI for election interference, such as the Deceptive Practices and Voter Intimidation Prevention Act.
• Pass the American Data Privacy Protection Act (ADPPA) to limit the collection and use of personal data for profiling and targeting.
• Implement an FTC Commercial Surveillance Rule that enforces data minimization and restricts secondary uses of personal information.
• Develop clearer legal standards for addressing AI-generated harms, including deepfakes, to ensure victims have access to effective remedies.

Conclusion

The report underscores the urgency of addressing the harms of generative AI through comprehensive legal and policy measures. It highlights the need for transparency, accountability, and fairness in AI development and use, calling for a proactive approach to mitigate risks and protect individuals and society from the negative consequences of AI-generated content.