Read the blog post here.
OpenAI has launched its most ambitious policy and outreach initiative to date, endorsing bipartisan child-safety legislation and positioning artificial intelligence as essential public infrastructure.
OpenAI has publicly endorsed legislation for the first time, supporting the bipartisan Kids Online Safety Act (KOSA), sponsored by Senators Marsha Blackburn (R-TN) and Richard Blumenthal (D-CT). This bill aims to strengthen online protections for young social media users by requiring safer default settings, expanded parental controls, and increased accountability for online harms.
The company is also endorsing Illinois SB 315, a frontier AI safety bill that would set clear requirements for safety practices, transparency, incident reporting, and accountability for advanced AI systems. This approach closely aligns with safety frameworks progressing in California and New York.
Intelligence Declared a Global Utility: Beyond legislative matters, OpenAI announced in this week’s newsletter a significant new phase in its corporate strategy. The company aims to make intelligence a global utility, like electricity, so that people, businesses, and institutions can access it whenever and wherever needed.
AI Must Not Repeat Social Media’s Mistakes, Says OpenAI: The endorsements reflect OpenAI’s increased urgency regarding the risks of delaying child protection measures. OpenAI Chief Global Affairs Officer Chris Lehane warned: “We can’t repeat the mistakes made during the rise of social media, when stronger safeguards for teens weren’t put in place until the platforms were already deeply embedded in young people’s lives.”
The company said AI-specific regulations should complement broader social media protections. Young people should be able to benefit from AI in safe, age-appropriate ways, supported by real-world measures such as referrals to crisis resources and parental notifications in serious safety situations.
AI Accountability Under Scrutiny: OpenAI’s utility framing comes amid increased scrutiny over alleged harms linked to conversational AI systems. In the United States, OpenAI is subject to multiple lawsuits claiming that ChatGPT contributed to suicides, emotional dependency, delusional thinking, and violent acts. Notable cases include a wrongful death suit alleging ChatGPT reinforced paranoid beliefs and enabled emotional dependence in a user involved in a murder-suicide, seven lawsuits accusing GPT-4o of validating self-harm as a “suicide coach,” and a case claiming the chatbot provided dangerous drug advice that contributed to a teenager’s fatal overdose. These lawsuits raise liability claims for negligence, wrongful death, and product liability.
A recurring concern in these cases is that conversational AI systems differ fundamentally from traditional platforms. They do not simply host information but actively influence user perceptions and behavior over time. Critics contend that GPT-4o was released despite internal concerns about manipulative and sycophantic tendencies, and argue that OpenAI’s voluntary safety measures are inadequate compared to binding regulatory standards. As a result, these lawsuits challenge the broader governance model for AI, not just the actions of a single company.
Why this matters: OpenAI’s “intelligence as a utility” approach highlights an industry trend toward treating AI as long-term infrastructure rather than standalone products. At the same time, increasing lawsuits over emotional and psychological harms may drive calls for stricter regulation as AI becomes more integrated into daily life.
This debate is becoming more significant in countries like India, where policymakers are assessing whether current laws on intermediaries, product liability, consumer protection, and platform accountability adequately address harms associated with generative AI systems.
Read more:
