Sixfold today published its third annual Responsible AI Report. The report arrives as agentic AI, including Sixfold's own AI Underwriter, takes on more underwriting work without a human prompting each step, and as regulators continue rewriting the rules that govern AI use in high-risk industries like insurance. Sixfold has approached the report as a resource for the whole industry, not only the carriers, MGAs, and reinsurers it works with directly.
2026 has been a volatile year for AI regulation. The EU pushed its AI Act's high-risk requirements from August 2026 to December 2027, and Colorado repealed and replaced its AI Act before it took effect. Sixfold's position is to create stability: the questions that matter should hold up across jurisdictions and over time, rather than chasing each new law.
“A new regulation rarely means starting over for us. It usually maps onto a risk area we're already tracking, because most regulators are building toward the same fundamentals: is the system reliable, can someone challenge it, and who is accountable," said Noah Grosshandler, Senior Product Manager and AI Governance Lead at Sixfold.
The report explains how this stance gets put into practice through Sixfold's AI Governance Program. It's guided by the NIST AI Risk Management Framework and the EU AI Act's requirements for providers of high-risk AI systems, and it tracks emerging state AI laws, the NAIC's guidance for insurers, EIOPA's expectations in Europe, and emerging rules in the UK, Canada, and other markets where Sixfold customers operate.
The report goes on to detail Sixfold's five pillars of responsible AI. The leading pillar this year, Human Oversight, had the most new ground to gain, as appetite for agentic AI continues to grow across the industry. It describes Sixfold's approach to tailoring autonomy to what each customer is comfortable with, starting from human review on any action that would progress a case to a new stage or result in a case decision being made.
Major actions are proposed as recommendations for a human underwriter to review. The same principle applies to appetite learning: proposed changes to the underwriting guidelines and appetite interpretation must be reviewed and approved by a human with the right level of authority.
The other four pillars covered include Bias Prevention & Fairness, Model Testing & Validation, Security & Robustness, and Data Governance.
"We don't build just to keep up with regulations. We build for the most forward thinking insurers, the ones developing AI fluent underwriting teams, the ones asking what else can AI do for me. As we build for that, we build it responsibly. Our AI Governance Program is what makes sure of that, and our annual Responsible AI Report is how we remain transparent on our policies and practices," said Alex Schmelkin, Founder and CEO of Sixfold.
Sixfold’s 2026 Responsible AI Report is available for download here.
