Who’s Accountable? Regulation, Trust, and Safety in the AI Era

Session Summary

NEW YORK, NEW YORK – SEPTEMBER 22: (L-R) Michael Loeb, Founder and Ceo Of Loeb.nyc, Jess Leinwand, Chief Legal Officer and General Counsel of UNICEF USA, Jenny Lay-Flurrie, Corporate Vice President and Head of the Trusted Technology Group at Microsoft, Rob Bonta, Attorney General of the State of California, and Hooman Shahidi, Managing Partner at Opus Bloom Capital, speak during the 2026 Concordia Annual Summit at Sheraton New York Times Square on September 22, 2026 in New York City. (Photo by Leigh Vogel/Getty Images for Concordia Summit)

Important
Quotations

“Child safety very specifically is an enterprise risk and a legal compliance obligation, not kind of an optional product design consideration.”
Jess Leinwand
“Hope is not a strategy and … we unequivocally have to come together, red or blue, Republicans and Democrats, private and public sector, because this is the fate of the future that’s in front of us.”
Hooman Shahidi
“We have the opportunity on the federal level to pick and choose, you know, the best policies and the best legislations that will come out of it… presumably, at the end, we’re going to get the best result.”
Michael Loeb
“We need to have responsible innovation. We can’t let innovation outpace accountability.”
The Hon. Rob Bonta
“At every pass and every stage, you have to be humanist in your approach, and we have to carry that forward.”
Jenny Lay-Flurrie

Key
Takeaways

  • Trust is the central currency of the AI era, and investment capital is currently moving faster than regulation can keep up. 
  • Innovation must not outpace accountability, and that existing criminal, civil rights, consumer protection, and health care laws already apply to AI, alongside new AI-specific laws. 
  • Core principles for responsible AI are protecting children, avoiding catastrophic risk, being transparent about when people are interacting with AI, honoring intellectual property and privacy, and preventing built-in discrimination. 
  • Trust has many facets, including privacy, human rights, dignity, accessibility, and security, and that AI can help measure performance but human experts must remain in the loop at every stage. 
  • After this summer’s lawsuits, child safety is now treated as an enterprise risk and a legal compliance obligation rather than an optional design consideration. 
  • On federal versus state regulation, a “both and” approach: the federal government sets a floor while states, cities, NGOs, and the private sector experiment and raise the ceiling. Even though the patchwork is painful for businesses. 
  • AI is advancing toward self-improving systems amid intense global competition, so regulation has to strike a careful balance between moving too slowly and leaving dangerous gaps. 
  • Government benefits systems are especially exposed: roughly $2 trillion is disbursed each year, an estimated $200 billion or more is stolen, and aging systems have little defense against AI-driven fraud. 
  • AI has lowered the barrier to nonconsensual intimate imagery, citing a figure of 1.42% of people affected in some countries, and pointed to the UK and EU as regulatory models worth learning from. 
  • Balance between children’s safety and children’s rights around the child’s best interest. Companies should choose the least restrictive way to reach each safety goal and design for context, since risks in the Global South differ from those in New York City. 
  • Regulatory cost does factor into startup investment decisions, and panel pointed to capital and people moving from New York and California to Texas and Florida. 

Action
Items

  • Conduct digital child impact assessments before products ship, modeled on the privacy impact assessments companies already use. 
  • Treat foreseeable risks to children as compliance obligations, with clear ownership and mitigation assigned across product, legal, and engineering teams. 
  • Build safety by design into every product, including age assurance and age-differentiated experiences. 
  • Keep qualified human experts in the loop to evaluate AI-generated data and outputs, especially on accessibility and safety. 
  • Clearly disclose to users when they are interacting with AI. 
  • Audit AI products against existing laws, including civil rights, consumer protection, privacy, and IP, and test for explicit or implicit discrimination. 
  • Modernize legacy government benefits systems and invest in AI-powered fraud detection before AI-driven fraud accelerates. 
  • Push for a federal regulatory floor that preserves state flexibility, and create channels for states, cities, NGOs, and industry to share what works. 
  • Study international approaches such as the UK’s and the EU’s, and build cross-border partnerships and public education to combat nonconsensual intimate imagery. 
  • Apply a “least restrictive means” test to child safety measures and tailor design to local context, including the Global South. 
  • Advise founders to factor regulatory compliance and jurisdiction into their business plans from the start.