Americans care about their privacy, but protecting it is hard. 94% of Americans are concerned about the privacy of their personal data, but struggle to protect it given arduous and lengthy processes designed to discourage them.
AI exposes your data in new ways. Not only will AI make identification cheaper and quicker at scale, it will likely continue to make use of new signals to further infer our characteristics or thoughts. One important development is recent progress in AI’s ability to associate brain waves with thoughts; further advances in AI and brain-scanning technology could bring remarkable mind-reading capabilities.
Continued AI development could contribute to further use of our data for manipulation. The practice is not new, but AI stands to increase both the effectiveness and frequency of attempts to influence us by using our data. Whether by leveraging our characteristics (age, education, medical history, etc) or our personalities (e.g. introvert vs extrovert), this capacity won’t just be used to get us to buy products, but will likely be expanded out for other manipulative means, like trying to influence our votes.
Our current system of privacy protection is failing. U.S laws largely follow the "notice-and-choice" approach, where companies inform people about their practices, and then people must consent to proceed. This practice is flawed because it gives people two unappealing choices: forfeit privacy rights, or forfeit access to the company's product, service, or website. A second problem is that, as AI’s predictive capabilities improve and further vision is granted into our lives, privacy violations will increasingly come from inferences models make about us rather than especially sensitive data we’ve given up. Voluntary steps could protect our privacy, but thus far companies have more often chosen to subvert privacy protections rather than protect them.
Comprehensive data protection is needed. The bipartisan American Privacy Rights Act (APRA) would be a huge step forward in safeguarding Americans’ privacy, requiring: transparency over how data is used, standards for data minimization, the option for consumers to access and delete their data, and more. Some changes could be made which would further strengthen the bill: government organizations should also be covered, opt-in consent for targeted ads would be an improvement over the option to opt-out. But even without these changes, this bill would greatly improve the current situation, and if passed would be a foundation upon which further privacy protections could be built. Americans are in support too, 80% approving of its major tenets.
Read the full report here.
What AI music generation will mean for our creativity, current and future musicians, and the art in general
Inspecting the claim that AI safety and US primacy are direct trade-offs
Policymakers and engineers should prioritize alignment innovation as AI rapidly develops