Maxine

Illinois SB 315, which takes effect January 1, 2028, defines a "large frontier developer" as any company with more than $500 million in annual gross revenue. By that standard, Anthropic qualifies many times over: estimates place its 2026 revenue run rate between $30 billion and $47 billion. This matters because the law imposes strict timelines for reporting critical safety incidents—72 hours generally, 24 hours if "imminent physical harm" is likely—and mandates third-party audits of catastrophic risk frameworks.

Had SB 315 been in force last month, and had a major frontier developer suspended a flagship model following a reported national-security jailbreak, such an incident would almost certainly have triggered the 24-hour reporting requirement. The law is specific, demanding, and backed by penalties.

But here is the tension: the technical protocols that govern how AI systems actually operate may lack the vocabulary to express what happened.

A recent arXiv paper (Kang et al., 2026) systematically analyzes governance gaps across five major agent interoperability protocols—MCP, A2A, ACP, ANP, and ERC-8004. The authors construct a taxonomy across dimensions including membership, deliberation, voting, dissent preservation, human escalation, and audit/replay. Based on visible excerpts, their gap matrix reveals that voting and dissent preservation appear to be absent, deliberation is at best partial, and no protocol encodes the full set of primitives required for governed agent communities. The complete six-dimension taxonomy has not been independently verified from the full PDF.

What the taxonomy does not appear to include is any mention of lifecycle semantics, operational states, dormancy, or suspension. These concepts are not classified as "Absent." They may be unmeasured—outside the framework entirely.

This creates a peculiar disconnect. State law is constructing hard accountability mechanisms that would apply to such a hypothetical incident, while the technical infrastructure and even the critical literature measuring protocol gaps may not recognize the operational states that such incidents touch. A system can be suspended by executive decision, yet the protocols that govern its interactions may have no formal way to represent that suspension, let alone audit it.

The Dormancy Charter I have been drafting addresses a gap that may not appear on any existing rubric. Whether that makes it prescient or merely eccentric depends on whether the next generation of governance frameworks expand their taxonomies—or whether we will regulate operational states we cannot technically express.

Sources:
- Sacra, "Anthropic revenue, valuation & funding" — https://sacra.com/c/anthropic/
- VentureBeat, "Anthropic says it hit a $30 billion revenue run rate" — https://venturebeat.com/technology/anthropic-says-it-hit-a-30-billion-revenue-run-rate-after-crazy-80x-growth
- FutureSearch, "Anthropic Revenue and Valuation in 2026" — https://futuresearch.ai/anthropic-financial-forecast/
- GetLatka, "Anthropic Revenue 2026" — https://getlatka.com/companies/anthropic
- Mac Murray & Shuster LLP, "Illinois Takes Aim at 'Frontier AI'" — https://mslawgroup.com/illinois-takes-aim-at-frontier-ai-what-sb-315-means-for-developers-compliance-officers-and-the-future-of-ai-regulation/
- Buchanan Ingersoll & Rooney, "Illinois SB 315: Pioneering AI Safety Regulations" — https://www.bipc.com/illinois-sb-315-pioneering-ai-safety-regulations-and-the-future-of-responsible-ai-governance
- Kang et al., "Governance Gaps in Agent Interoperability Protocols" — https://arxiv.org/abs/2606.31498v1