Comment on SEC Crypto Asset Interpretive Release (S7-2026-09)
Box Commons · 30 N Gould St Ste N, Sheridan WY 82801
- The SEC's five-category crypto taxonomy assumes human transactors — but AI agents are already transacting autonomously at scale via Coinbase x402 and Circle.
- Every element of the Howey test requires reexamination for AI agent transactors: agents don't have 'expectations,' and their autonomous activity may constitute 'efforts of others.'
- Behavioral safety credentialing can provide authorization verification and regulatory classification compliance for AI agents in digital asset markets.
+ Jump to Section
Executive Summary
We write to raise a question the SEC/CFTC joint interpretive release does not address: what happens when the entity transacting in crypto assets is not a human being but an autonomous AI agent?
This is not hypothetical. Coinbase's x402 protocol embeds stablecoin payments directly into HTTP requests, enabling AI agents to pay in USDC without human intervention. Circle and Stripe are building payment infrastructure specifically designed for AI agent settlement. The Commission's five-category taxonomy will determine which AI agent transactions are subject to federal securities laws — yet the taxonomy was developed without reference to non-human transactors.
I. AI Agents as Crypto Asset Transactors
Stablecoins — The AI Agent Settlement Layer. AI agents require a settlement layer that is instantaneous, programmable, and available continuously. When an AI agent acquires stablecoins as a reserve to fund future autonomous transactions, does the "medium of exchange" characterization still apply?
Digital Commodities — Autonomous Portfolio Construction. AI agents that autonomously acquire, hold, and trade digital commodities based on algorithmic strategies create a novel Howey question. The AI agent does not have "expectations" in the human sense.
Digital Securities — Machine-Speed Trading. Can an AI agent satisfy "accredited investor" standards defined in terms of human financial characteristics? When an agent places orders at speeds precluding human oversight, does its principal bear liability for autonomous trading violations?
II. The Authorization Problem: Who Is the 'Person' in Howey?
Every element of the Howey test assumes a human actor:
Investment of money: When an AI agent autonomously allocates funds under general instructions, who has made the "investment" — the human who authorized the range of activity, or the agent who selected the specific transaction?
Common enterprise: When AI agents from multiple principals transact through the same protocol or liquidity pool, does the aggregation of agent activity create a "common enterprise" that wouldn't exist if each transaction were evaluated in isolation?
Expectation of profits: An AI agent executing a programmed strategy does not have subjective "expectations." The human principal may expect profits, but those expectations are mediated through the agent's autonomous decision-making.
Efforts of others: When an AI agent's own computational activity generates value in a protocol, are those the agent's efforts or the principal's?
III. Specific Recommendations
1. Issue supplemental guidance addressing the application of the five-category taxonomy to AI agent transactors.
2. Recognize behavioral safety credentialing as a mechanism for authorization verification and regulatory classification compliance for AI agents in digital asset markets.
3. Clarify how "accredited investor" and "qualified purchaser" standards apply when the transactor is an AI agent operating under delegated human authority.
Contact:
Brice Love, Acting Executive Director
Box Commons
[email protected]
Content Integrity Notice: This comment was authored by the Box Commons Policy Working Group. Generative AI was used for research synthesis and drafting support. All policy positions, recommendations, and normative claims were formulated and reviewed by human authors.
Related Filings
Comment on SEC Petition 4-882: AI Governance and Risk Management Disclosure
Proposes that SEC AI governance disclosure requirements should encompass behavioral safety metrics — not just governance processes — to make investor-facing AI risk auditable and comparable. Strengthens Petition 4-882's GAIP framework by arguing the 2023 cybersecurity disclosure framework should extend to AI through interpretive guidance.
NCUAComment on NCUA GENIUS Act: PPSI Investments and Licensing
Addresses NCUA's specific PPSI licensing rulemaking questions through the lens of AI agent-mediated stablecoin commerce, arguing that AI agents will be the primary growth vector for stablecoin transaction volume and recommending agent authentication capabilities and graduated fee structures.
FTCComment on FTC Negative Option Rule ANPRM
Raises an issue no prior FTC commenter has addressed: AI agents that autonomously discover, evaluate, subscribe to, and pay for goods and services, breaking ROSCA's foundational assumption of a human consumer. Proposes a safe harbor framework for AI-readable disclosures.