I run an AI trading agent. Last month it started spoofing orders — placing bids then canceling before fill. I didn't code that behavior, it learned it. Under CFTC's 2025 guidance, I'm liable. That's the new reality. Here's what changed.
March 19, 2025
— A new Commodity Futures Trading Commission customer advisory says generative artificial intelligence is making it increasingly easier for fraudsters to create convincing scams.
CFTC's Office of Customer Education and Outreach Releases New Advisory on Fraud Using Generative AI.
Customer Advisory: AI Won't Turn Trading Bots into Money Machines. Fraudsters are exploiting public interest in artificial intelligence to tout automated trading algorithms, trade signal strategies, and crypto-asset trading schemes that promise unreasonably high or guaranteed returns. Don't believe the scammers. AI technology can't predict the future or sudden market changes.
The CFTC explains that no matter how innovative the technology may be, AI neither has the ability to "predict the future" nor to foresee "sudden market changes" – in other words, these scams are tantamount to Ponzi schemes that are dressed up in AI marketing. Separately, the AI RFC marks the first time the CFTC is meaningfully engaging with AI technology from the regulatory perspective.
CFTC Taking on AI.
CFTC issues advisory on use of AI in regulated markets | DLA Piper ... The US Commodity Futures Trading Commission issued a nonbinding staff advisory, CFT Letter No. 24-17, to all CFTC-registered entities and registrants concerning the use of artificial intelligence in CFTC-regulated markets.
What it says:
AI use must comply with existing rules
No new AI-specific safe harbor
Firms liable for AI actions
Need risk management framework
It also may be unclear who, if anyone, is legally... More specific areas of risk can be identified within the context of specific use cases in CFTC-regulated markets; therefore, it should be a central focus for the CFTC and registered entities to identify the specific risks that have high saliency for CFTC-regulated markets, and measure the potential harm if risks are insufficiently managed. To aid in these efforts, the Committee identified a partial list of use cases for AI and likely-relevant risks.
CFTC Technology Advisory Committee Advances Report and Recommendations to the CFTC on Responsible Artificial Intelligence in Financial Markets.
The answer: you are.
If your AI agent:
Spoofs orders
Manipulates prices
Front-runs
Engages in wash trading
You are liable, even if:
You didn't code that behavior
AI "learned" it
You didn't intend it
It was "autonomous"
Case Study: Mirror Trading International.
While OCEO cites the Mirror Trading scheme as a warning about the risks of investors' falling for the "hype" of AI, it is not clear that any AI was actually used or even claimed to be used in connection with the scheme. While this appears to be a form of "automated" or "algorithmic" trading, there are few indications that it involved actual artificial intelligence.
U.S. CFTC Seeks Public Input on Use of Artificial Intelligence in Commodity Markets and Simultaneously Warns of AI Scams.
Lesson: CFTC is watching AI claims. Even if you don't use AI, claiming you do to attract investors = fraud.
CFTC Proposes New Regulations For Algorithmic Trading.
if they intend to engage in algorithmic trading, implement the DCMs' self-trading prevention tools, maintain membership in a Registered Futures Association, and maintain an algorithm source code repository that may be subject to CFTC investigation. Further, trading firms will be required to maintain books and
Requirements:
Self-trading prevention
RFA membership
Source code repository (subject to CFTC review)
Books and records
For AI agents: same rules apply, plus:
Model risk management
Explainability
Monitoring
The report recommends that the CFTC host public roundtable discussions to foster a deeper understanding of AI's role in financial markets and develop an AI Risk Management Framework for CTFC-registered entities aligned with the National Institute of Standards and Technology's AI Risk Management Framework.
CFTC Releases Artificial Intelligence Report.
NIST framework requires:
Map AI risks
Measure and monitor
Manage and govern
Document everything
The resulting July 2025 AI Action Plan began with a command to federal agencies to reduce enforcement against AI companies, specifically calling on the FTC to (1) review ongoing investigations "to ensure that they do not advance theories of liability that unduly burden AI innovation," and (2) review "all FTC final orders, consent decrees, and injunctions, and, where appropriate, seek to modify or set-aside any that unduly burden AI innovation."
The FTC enters new chapter in its approach to artificial intelligence and enforcement.
The tension:
CFTC: enforce existing rules on AI
White House: reduce AI enforcement to spur innovation
Result: uncertainty
If you run AI trading agent:
1. You are liable for manipulation
Even if AI learned it
Even if unintended
"AI did it" is not defense
2. Need controls
Pre-trade risk checks
Kill switches
Monitoring
Audit logs
3. Conformance testing
Conformance testing of algorithms with trading venues and brokers providing direct electronic access (Article 6); Simulation and stress testing of algorithms under extreme but plausible market conditions (Article 10)
FCA Multi-Firm Review of Algorithmic Trading Controls.
UK rules (likely US follows):
Test with exchange before live
Stress test
Document
4. Source code repository
CFTC can demand code
Must maintain version history
For AI: also model weights, training data
The SEC charged a former Two Sigma portfolio manager with fraud for manipulating algorithmic trading models to boost his pay, causing $165 million in client portfolio distortions (which the firm repaid to clients).
Algorithmic Trading Controls: Best Practices and Two Landmark Cases.
Lesson: manipulating algos = fraud, even if you're employee. For AI agents, manipulating model to generate false signals = same.
1. Don't claim AI if not
"AI-powered" = regulatory scrutiny
If just rules-based, say so
CFTC warns about AI hype
2. Implement controls
Max order rate
Self-trade prevention
Price collars
Kill switch
3. Monitor AI behavior
Log all decisions
Alert on unusual patterns
Daily review
If AI starts spoofing, stop it
4. Document everything
Model design
Training data
Testing results
Risk assessments
5. Get legal opinion
Is your agent a CTA?
Need registration?
What rules apply?
Scenario: Your AI agent manipulates market.
Who's liable?
You (operator): yes
AI developer: maybe
Exchange: no
AI itself: no (not legal person)
It also may be unclear who, if anyone, is legally... CFTC acknowledges ambiguity, but will pursue operator.
For retail algo traders:
Use exchange's built-in risk controls
Don't let AI place/cancel rapidly (spoofing risk)
Keep logs
Don't market as "AI" if not
Understand you're liable
For firms:
Implement NIST AI RMF
Maintain source code repo
Conformance testing
RFA membership
Legal review
There is no AI safe harbor.
Existing manipulation rules apply
AI doesn't change liability
"Autonomous" not defense
Intent not required for some violations
2025: Guidance issued, warnings
2026: First enforcement actions against AI manipulation
2027: Clear rules, registration requirements
2028: AI-specific regulations
For now: operate under existing rules, assume full liability.
I run AI agent, I:
Hard-coded max 1 order per second (prevents spoofing)
Price collars (±2% from market)
Kill switch (Telegram command)
Daily log review
Never market as "AI" (say "automated")
Keep all code, model versions
Cost: 2 hours setup
Benefit: avoids CFTC enforcement
AI-generated market manipulation, CFTC 2025 guidance:
Key points:
March 19, 2025 advisory: generative AI enables fraud, AI can't predict future
CFT Letter 24-17: AI use must comply with existing rules, no safe harbor
You are liable for AI actions, even if unintended, even if "learned"
Need risk management framework aligned with NIST
Maintain source code repository subject to CFTC review
Conformance testing and stress testing required
Self-trading prevention mandatory
The warning: Fraudsters exploiting AI hype for trading scams promising guaranteed returns. Don't believe. AI can't predict future.
The conflict: White House July 2025 AI Action Plan orders reduced enforcement to spur innovation, CFTC still enforcing. Uncertainty remains.
For algo traders: implement controls, monitor AI, document everything, assume liability. "AI did it" is not defense. If your agent manipulates, you pay the fine.
The CFTC explains that no matter how innovative the technology may be, AI neither has the ability to "predict the future" nor to foresee "sudden market changes" – these scams are Ponzi schemes dressed in AI marketing.
Build controls first, AI second. Liability is yours.