New York\u00a0\u2014\u00a0\n <\/p>\n
\n Since independent vehicle crash testing began in the mid-1990s, automakers have been incentivized to make safety changes that have saved thousands of lives each year.\n <\/p>\n
\n Now, a new group is hoping to take a similar approach to artificial intelligence.\n <\/p>\n
\n Nonprofit media watchdog Common Sense Media is launching the Youth AI Safety Institute, an industry-backed, independent research and testing lab to study the risks AI tools may pose to children and teens. It will aim to provide information to parents and families about various AI tools and set safety benchmarks for tech firms.\n <\/p>\n
\n AI companies are locked in a race to build the most powerful, widely used models, and that sometimes means speed is prioritized over safety testing. Because AI tools are complex systems with a range of different uses, ranking their safety will likely be far trickier than judging how a car responds in a crash.\n <\/p>\n
\n But Common Sense Media and the board of top AI, education and health leaders it recruited to oversee the Youth AI Safety Institute believe that solely relying on AI firms to self-police on safety isn\u2019t enough to protect young people. Existing third-party AI safety organizations largely focus on societal-level and existential risks, such as job loss or even human extinction, rather than consumer-friendly safety ratings aimed at everyday use.\n <\/p>\n
\n The goal is for the public spotlight and third-party standards to spark what Common Sense Media CEO James Steyer called a \u201crace to the top\u201d for tech firms to make safety fixes to improve their standing.\n <\/p>\n
\n Leading AI firms invest in safety research to \u201cmake their models as good as they possibly can, but there\u2019s no independent measure of that,\u201d John Giannandrea, Apple\u2019s former AI strategy chief who joined the institute\u2019s advisory board, told CNN. \u201cWe don\u2019t really know which models are more appropriate for kids at a certain age than others, and I think the only real way to do that is to have an independent set of public standards.\u201d\n <\/p>\n
\n The launch comes as multiple families have sued<\/a> AI companies alleging that chatbots encouraged their children\u2019s suicides. A recent CNN investigation found<\/a> that AI chatbots advised teen test accounts on how to commit violence. Grok, xAI\u2019s chatbot, came under fire<\/a> earlier this year for sharing sexualized images of women and children in response to users\u2019 \u201cdigital undressing\u201d prompts. And growing AI adoption in classrooms has raised questions about whether the technology could stunt learning<\/a>.\n <\/p>\n \n \u201cI think many parents and educators and citizens feel we\u2019re at a catastrophic moment as AI is reshaping the lives of children and families and schools and, quite frankly, all of society,\u201d Steyer told CNN exclusively ahead of announcing the group on Tuesday.\n <\/p>\n \n The Institute will start with a $20 million annual budget, backed by OpenAI, Anthropic and Pinterest, as well as the Walton Family Foundation, Goldman Sachs Managing Director Gene Sykes and other philanthropists. Funders will have no say in the group\u2019s operation or research, according to Common Sense.\n <\/p>\n \n The group\u2019s advisory board will also include Mehran Sahami, chair of Stanford University School of Engineering\u2019s computer science department; Dr. Jenny Radesky, director of University of Michigan Medical School\u2019s developmental behavioral pediatrics division; and Dr. Nadine Burke Harris, who served as California\u2019s first-ever surgeon general \u2014 bringing together expertise in research, standards setting and tech product development.\n <\/p>\n \n The institute will \u201cred team\u201d leading AI models and products used by young people \u2014 stress testing them to identify potential risks or shortcomings in safety guardrails. It will then publish research as consumer-friendly guides for the public and develop AI youth safety standards, or benchmarks, that tech companies could use to develop or improve their products. It plans to release research starting this month.\n <\/p>\n \n AI companies already use such benchmarks to measure and compare their performance across other metrics. The group hopes the public pressure, as well as its industry connections, will encourage AI companies to incorporate the standards into their development and testing \u2014 and make safety changes to improve their standings.\n <\/p>\n \n \u201cBenchmarks are really the lifeblood of how people measure and how we know all this investment is resulting in higher quality models,\u201d Giannandrea said. \u201cWhat we need is a benchmark for harm, and specifically for child harm.\u201d\n <\/p>\n \n Among the challenges for researchers is the pace of AI development. Unlike physical products that are released on a regular cadence and may not change much once they hit the market, AI models often gain new update capabilities \u2014 and thus potentially new risks \u2014 on a weekly or monthly basis.\n <\/p>\n \n Establishing the Youth AI Safety Institute as a separate group will enable even more frequent, robust research to keep up with the rapid advancement of AI models, Steyer said.\n <\/p>\n \n Common Sense Media is widely used by parents and educators for its ratings of movies, video games and other online platforms; the organization says its platforms have 150 million monthly users. And it has already been studying AI-related risks. Last year, it warned<\/a> that AI companion apps pose \u201cunacceptable risks\u201d to young people.\n <\/p>\n \n It has also published risk assessments<\/a> of AI tools such as OpenAI\u2019s ChatGPT, MetaAI and Grok. Those reports rank the tools on a scale from \u201cminimal risk\u201d to \u201cunacceptable\u201d on measures including kids\u2019 safety, data use and trustworthiness, and provide examples of where the tools fall short.\n <\/p>\n \n The Youth AI Safety Institute wants to avoid a repeat of the social media era\u2019s safety pitfalls. It took years before whistleblowers, investigative reports and lawsuits revealed the full scope of the risks social apps pose to young people. Earlier this year, a California jury found Meta and YouTube liable<\/a> for knowingly addicting and harming a young woman in a landmark decision, decades after the platforms launched.\n <\/p>\n