The Growing Problem of Deceptive AI-Generated Content Online

Artificial intelligence has made incredible strides in the past few years, with AI chatbots like ChatGPT and image generators like DALL-E and Midjourney showcasing the technology‘s potential. However, the rise of AI has also enabled the mass production of fake news, scams, deepfakes, and other misleading content online.

A 2024 survey by cybersecurity firm CyberVista found that a staggering 85% of internet users reported being fooled by AI-generated content at least once in the past year, up from 77% in a similar 2023 survey by All About Cookies. As generative AI continues advancing at a breakneck pace, experts predict this number will only grow higher.

Confronting the Legal and Ethical Challenges of AI Content

The proliferation of deceptive AI content raises a host of thorny legal and ethical issues that policymakers are struggling to address:

• Labeling AI content: One proposal is to legally require all AI-generated content to be clearly labeled as such. However, no foolproof technical method currently exists for detecting AI content, especially as the technology grows more sophisticated. Enforcing labeling requirements would be extremely difficult.
• Copyright infringement: Many AI systems are trained on copyrighted works like books, articles, and images scraped from the web without permission. When these AI tools create new content, it often mimics or builds upon these copyrighted works in ways that could be considered infringement. Artists and writers have filed lawsuits alleging that AI art and text generators violate their intellectual property rights.
• Defamation and libel: AI language models can easily generate fake news stories, reviews, and social media posts that include false and defamatory statements about individuals or organizations. However, holding AI companies liable for defamation is legally tricky.
• Nonconsensual deepfakes: AI allows users to easily create fake pornographic videos with celebrity likenesses, enabling a new form of sexual exploitation. Most experts agree that stronger legal protections are needed.
• AI and national security: The same AI tools that can generate fake news articles can also create convincing impersonations of world leaders, potentially increasing risks of military conflict and political turmoil. Policymakers fear AI could supercharge foreign disinformation campaigns.

Current laws are not well-equipped to handle these novel issues. Congress is considering the bipartisan AI Content Generation Disclosure Act, which would require AI-created content to be labeled and ban nonconsensual deepfakes. However, experts say much more comprehensive legislation is needed to address the full scope of challenges posed by AI content.

The Ethics of AI Training Data

Another key issue is the personal data used to train AI systems. Most people surveyed by CyberVista (92%) agreed there should be stronger protections against companies using personal information to train AI without consent. Only allowing opt-in data collection was the most popular option.

However, some argue that restricting AI training data could hamper innovation and make AI systems less useful and inclusive. For example, limitations on medical data could impede development of AI diagnostic tools. There are also concerns that AI trained primarily on data from opt-in users — who tend to be more affluent and tech-savvy — could produce biased and discriminatory results.

Policymakers must strike a delicate balance between protecting individual privacy rights and realizing AI‘s societal benefits. Experts recommend:

• Requiring companies to disclose in clear terms how personal data may be used for AI training and allow users to opt out.
• Prohibiting use of sensitive categories of personal information, such as medical records and biometric data, without affirmative consent.
• Mandating strict security protocols for AI training data to prevent breaches.
• Encouraging development of privacy-preserving AI techniques that don‘t require collecting personal data.

Who Is Liable for AI‘s Misdeeds?

When AI tools are used to break the law or cause harm, it‘s often unclear who should be held responsible. Is it the user who generated the illegal content, the company that created the AI tool, or the website or platform that published the content?

In the CyberVista survey, 72% of respondents said the individual users who generate illegal AI content should be held liable. However, many experts argue that AI companies must also be held accountable, especially if their tools are specifically designed for malicious purposes or if they fail to implement reasonable safeguards against abuse.

Website operators could also face liability if they knowingly allow illegal AI content to proliferate on their platforms. Section 230 of the Communications Decency Act currently gives websites broad immunity for user-generated content. But some advocates argue Section 230 should be reformed to require more content moderation.

Ultimately, a combination of user, company, and platform liability is likely needed to effectively combat harmful AI content. Clear legal standards must be developed for different scenarios, guided by the principle that those who are best positioned to prevent harm should bear primary responsibility.

Staying Safe in an AI-Generated World

As online fakery grows more sophisticated, internet users must stay vigilant to avoid being deceived. Some tips:

• Use browser extensions that detect and label potential deepfakes and AI-generated text.
• Cross-reference suspicious information with authoritative sources before sharing or acting on it.
• Be wary of sensational headlines and too-good-to-be-true offers.
• Check personal information on data broker sites and opt out to reduce risk of AI mimicry.
• Support AI literacy education initiatives to help more people spot misleading content.

With the right mix of technological solutions, regulatory oversight, and public awareness, we can harness AI‘s benefits while mitigating its potential for harm. But the challenges posed by AI-generated content will only continue to grow. Tackling this issue is critical for the future of online trust and democracy.