Mr. Sanctions

Category: AI & Other Tech

  • Can you be worse than terse?

    My Business Writing professor said you couldn’t, but I don’t know. I asked Gemini to produce a summarized version of the OFAC Enforcement Guidelines, and this is what I got back:

    This document acts as the “Sentencing Guidelines” for violating U.S. economic sanctions (OFAC rules). If you or your company trade with a banned country, terrorist, or sanctioned entity, this guide dictates your punishment.

    Here is the condensed version of how it works:

    1. The Spectrum of Punishment

    When you break the rules, the outcome ranges from a slap on the wrist to a major blow:

    • No Action: They drop the case.
    • Cautionary Letter: A formal warning (“Don’t do it again”).
    • Finding of Violation: A permanent mark on your record, but no money is owed.
    • Civil Monetary Penalty: You are fined cash.

    2. The “Scorecard” for Fines

    If they decide to fine you, they don’t pick a random number. They use a formula:

    • Base Fine: They start with a standard amount based on the dollar value of the illegal transaction.
    • Adjustments: They raise or lower that amount based on General Factors:
      • Fine Increases: If you did it on purpose (willful), were reckless, tried to hide it, or have a history of breaking rules.
      • Fine Decreases: If you cooperated, had a compliance program in place to stop it (but it failed), and took immediate steps to fix the error.

    3. The “Golden Rule” (Voluntary Self-Disclosure)

    This is the most critical part of the document. If you find your own mistake and tell the government about it before they catch you, your base penalty is automatically cut in half (50% off).

    The Bottom Line:

    OFAC treats honest mistakes much differently than reckless law-breaking. If you mess up: Confess early, cooperate fully, and fix the problem to minimize your penalty.

    So, I thought some important detail was missing, so I modified the prompt a few times:

    Summarize https://www.ecfr.gov/current/title-31/subtitle-B/chapter-V/part-501/appendix-Appendix%20A%20to%20Part%20501 in layperson-friendly terms. Include discussion of the penalty process, the General Factors which go into the calculation of the penalty, including any subcategories under each General Factor, and the Penalty Grid, which is based on egregiousness of the conduct and voluntary self-disclosure. Include the entire Applicable Schedule Amount table, and the current statutory maximum fines under TWEA, IEEPA, and the Kingping Act. Include a short introduction, and a short section of takeaways.

    and got back the following:

    Here is a summary of Appendix A to Part 501 of Title 31 (Economic Sanctions Enforcement Guidelines), broken down into layperson-friendly terms.

    Introduction

    This document is the official “rulebook” used by the Office of Foreign Assets Control (OFAC)—the agency that enforces U.S. sanctions (like those against Iran, Russia, or terrorist groups). It explains exactly how the government decides on the punishment when an individual or company violates these sanctions. It is designed to make penalties predictable, consistent, and transparent.

    The Penalty Process

    When OFAC suspects a violation, they follow a formal legal process. They do not simply send a bill; you have opportunities to explain yourself.

    1. Investigation: OFAC gathers evidence, often by issuing a subpoena (a legal demand for documents) or by reviewing a voluntary self-disclosure you submitted.
    2. Initial Determination: OFAC reviews the facts and chooses an outcome:
      • No Action: The case is dropped.
      • Cautionary Letter: A formal warning. No fine is issued, but you are told to fix your compliance mistakes.
      • Finding of Violation: A formal record that you broke the law, but without a monetary fine.
      • Civil Monetary Penalty: A cash fine.
    3. Pre-Penalty Notice: If they choose to fine you, they send this notice first. It explains what you did wrong and the proposed fine amount.
    4. Response: You usually have 30 days to reply. You can argue that you are innocent or admit the mistake and argue for a lower fine based on the “General Factors” (see below).
    5. Final Penalty Notice: After reading your response, OFAC issues the final fine.
    6. Settlement: At almost any stage, you can negotiate a settlement to resolve the case (usually for a lower amount) without going through the full legal fight.

    General Factors (The “Scorecard”)

    OFAC uses these factors to judge how “bad” the violation was. These are the levers that raise or lower your penalty.

    A. Willful or Reckless Violation of Law (Intent)

    • Willfulness: Did you know it was illegal and do it anyway?
    • Recklessness: Were you careless? Did you ignore red flags?
    • Concealment: Did you try to hide the evidence?
    • Pattern of Conduct: Was this a one-time accident or a habit?
    • Management Involvement: Did senior managers know?

    B. Awareness of Conduct (Knowledge)

    • Actual Knowledge: Did you know the transaction was happening?
    • Reason to Know: Should you have known? Was the info readily available?

    C. Harm to Sanctions Program Objectives (Impact)

    • Economic Benefit: Did money or goods actually reach the sanctioned target?
    • Implications for U.S. Policy: Did your action embarrass the U.S. or help an enemy?
    • License Eligibility: If you had just asked for permission first, would OFAC have said “yes”? (If so, the harm is lower).

    D. Individual Characteristics

    • Commercial Sophistication: Are you a large, global bank (held to a high standard) or a small local business?
    • Size & Financial Condition: Can you afford the fine?
    • Sanctions History: Have you been in trouble with OFAC in the last 5 years?

    E. Compliance Program

    • Did you have an internal system (software/rules) to prevent this?
    • Was that system appropriate for your company’s size?

    F. Remedial Response (Fixing It)

    • Did you stop the violation immediately?
    • Did you investigate yourself to find the root cause?
    • Did you implement new controls to ensure it never happens again?

    G. Cooperation with OFAC

    • Voluntary Self-Disclosure: Did you turn yourself in? (This is the most critical factor).
    • Responsiveness: Did you answer OFAC’s questions quickly and honestly?
    • Tolling Agreements: Did you agree to extend the statute of limitations to give OFAC more time?

    H. Timing: How long ago did this happen?

    I. Other Enforcement: Did another agency (like a banking regulator) already fine you?

    J. Future Compliance: Will a fine here serve as a good deterrent for others?

    The Penalty Grid

    OFAC calculates the “Base Penalty” using a matrix based on two questions:

    1. Was the conduct Egregious (severe/willful)?
    2. Did you Voluntarily Self-Disclose (confess)?
    ScenarioDefinitionBase Penalty Calculation
    1. Non-Egregious + DisclosedYou made a mistake and told OFAC yourself.50% of the Transaction Value (Capped at $188,850 per violation).*
    2. Non-Egregious + No DisclosureYou made a mistake, but OFAC caught you.The “Applicable Schedule Amount” (See table below).
    3. Egregious + DisclosedYou did it on purpose, but you confessed.50% of the Statutory Maximum.
    4. Egregious + No DisclosureYou did it on purpose and OFAC caught you.The Statutory Maximum.

    *Note: For very small violations in Scenario 1, the penalty is capped at much lower amounts based on the schedule.

    Applicable Schedule Amount Table

    This table is used for Non-Egregious cases (mistakes) where you did not voluntarily disclose. The fine is based on the value of the illegal transaction.

    Transaction ValueApplicable Schedule Amount (Base Penalty)
    Less than $1,000$1,000
    $1,000 to $10,000$10,000
    $10,000 to $25,000$25,000
    $25,000 to $50,000$50,000
    $50,000 to $100,000$100,000
    $100,000 to $200,000$200,000
    $200,000 or moreThe Statutory Maximum (See below)

    Current Statutory Maximum Fines (2025)

    For “Egregious” cases (or high-value non-egregious cases), the penalty is the “Statutory Maximum.” These amounts are adjusted annually for inflation. As of January 15, 2025, the maximums per violation are:

    • IEEPA (International Emergency Economic Powers Act): The greater of $377,700 or twice the value of the transaction. (This covers most sanctions programs like Iran, Russia, and Terrorism).
    • TWEA (Trading With the Enemy Act): $111,308. (Primarily covers Cuba).
    • Kingpin Act (Foreign Narcotics Kingpin Designation Act): $1,876,699.

    Takeaways

    • Honesty Saves Money: Voluntarily self-disclosing a violation automatically cuts the base penalty calculation in half.
    • Intent is Expensive: The difference between “Egregious” (reckless) and “Non-Egregious” (mistake) is massive—often the difference between a $10,000 fine and a multi-million dollar fine.
    • Compliance is a Shield: Even if you fail, having a compliance program in place lowers your penalty under the “General Factors.”
    • Cooperate: Hiding documents or delaying the investigation is considered “Concealment” and will dramatically increase your fine.

    Depending on what level of detail you need for your role, you may be able to deal with the ultra-condensed version. But it’s a cautionary tale about asking AI tools to summarize without knowing the source material and knowing what you want to see come back. And I suspect the Goldilocks result is really a combination of the two responses… but that’s just me.

    What do you think?

    —

    , , ,

  • Last call for questions for OFSI/FCDO Webinar

    Final Call for Questions: Moving to a single list for UK sanctions designations – WebinarOn 17 December, FCDO and OFSI will be holding a webinar on the move to a single list for UK sanctions designations. In this webinar we will be discussing the change taking place on the 28 January 2026, an explanation of what improvements are being made to the UK Sanctions List and its search tool and steps to take to ensure you are prepared.

    Webinar:
    Wednesday 17 December
    11:00 – 12:00

    OnlinePre-submitted questions: if you wish to submit a question in advance, please email ofsiengagement@hmtreasury.gov.uk by 09:00 on Monday 15 December 2025.
    Register Here

    —

    , ,

  • Burning Questions: Was that oil tanker seizure legal?

    So, I wanted to see what Gemini thought of yesterday’s activities… So I asked (notice the guardrails):

    Assuming all the allegations by the US Department of Defense and the Trump Administration are true, given all the published, in-force regulation, guidance, legislation and Frequently-Asked Questions, etc. published by the US government in relation to Venezuela and Iran, was yesterday’s seizure of an oil tanker off the coast of Venezuela legal? Justify your decision and produce sources from US government websites.

    and this is what I got back:

    Based on the reports of the seizure of the oil tanker “The Skipper” (formerly Adisa) on December 10, 2025, and assuming all allegations by the Department of Defense and the Trump Administration are true, the seizure is legal under United States law.

    The legality is grounded in a combination of Executive Orders (sanctions), federal statutes (forfeiture and anti-terrorism), and judicial process (seizure warrant).1

    Legal Justification

    1. Violation of Sanctions & Material Support to Terrorism2

    The US government alleges the tanker was transporting oil from Venezuela and Iran and supporting the Islamic Revolutionary Guard Corps (IRGC), which is designated as a Foreign Terrorist Organization (FTO).3

    • Iran Sanctions: Under the International Emergency Economic Powers Act (IEEPA) and Executive Order 13224 (Counter Terrorism), the US prohibits transactions that provide material support to designated terrorist organizations (like the IRGC).4 The transport of Iranian oil is considered a mechanism for funding the IRGC’s Qods Force.
    • Venezuela Sanctions: Executive Order 13850 and Executive Order 13884 block the property of Petróleos de Venezuela, S.A. (PdVSA) and prohibit US persons or those within US jurisdiction from dealing in Venezuelan oil.
    • Secondary Sanctions: Even for non-US ships, providing “material support” or goods/services to a designated FTO or sanctioned entity can trigger US jurisdiction under extraterritorial sanctions provisions.

    2. Civil Forfeiture Authority

    The seizure was reportedly conducted pursuant to a seizure warrant issued by a US federal court.

    • 18 U.S.C. § 981(a)(1)(C) and (G): These statutes authorize the civil forfeiture of any property, real or personal, which constitutes or is derived from proceeds traceable to a violation of IEEPA, or assets of any foreign terrorist organization engaged in planning or perpetrating acts of terrorism against the US.
    • Because the oil and the vessel constitute “property” used to facilitate these violations or are proceeds of the violations, they are subject to forfeiture to the US government.

    3. Maritime Law Enforcement

    • 14 U.S.C. § 522: This statute grants the US Coast Guard (which participated in the operation) the authority to make inquiries, examinations, inspections, searches, seizures, and arrests upon the high seas and waters over which the United States has jurisdiction, for the prevention, detection, and suppression of violations of laws of the United States.
    • Jurisdiction over the Vessel: While the vessel was reportedly “off the coast of Venezuela” (likely international waters), the US often establishes jurisdiction if:
      • The vessel is stateless (not properly registered).
      • The flag state (in this case, reportedly Guyana) consents to the US boarding and seizure.
      • The vessel is deemed “property” subject to a US court’s forfeiture warrant.

    Sources from US Government Websites

    • Authority for Sanctions (Iran & Counter-Terrorism):
    • Authority for Sanctions (Venezuela):
    • Forfeiture and Seizure Authority:
      • Asset Forfeiture Program (Department of Justice): Explains the use of forfeiture to disrupt criminal activity, including terrorist financing. Justice.gov – Asset Forfeiture Program
      • (Precedent) 2020 Seizure of Iranian Fuel: The Department of Justice previously successfully executed similar seizure warrants for Iranian fuel bound for Venezuela aboard the vessels Bella, Bering, Pandi, and Luna, citing the same legal authorities (IEEPA, forfeiture for FTO support).5 Justice.gov – 2020 Press Release

    And the list of sources included The Washington Post, the U.S. Department of Justice website, Al Jazeera and Wikipedia (for what that’s worth)…

    Would welcome someone more expert than me, and with no axe to grind, giving their opinion

    —

    , ,

  • New OSFI AI Workshop

    FIFAI II: A Collaborative Approach to AI Threats, Opportunities, and Best Practices, Workshop 3 – AI and Financial Stability

    Here is a section-by-section summary in layperson terms.

    1. Introduction & Context

    AI is a big deal—it can do great things but also cause major problems. Financial leaders need to stop being scared and waiting for permission; instead, they should innovate responsibly. The goal of this workshop was to figure out how AI makes the financial system less stable and how to fix that.

    2. Three Avenues of Risk

    The workshop participants identified three main ways (or “avenues”) that AI could threaten the stability of the financial system.

    • Avenue 1 (Internal): Risks coming from banks and insurers using AI inside their own companies.
    • Avenue 2 (External): Risks arising when people or groups outside the financial system use AI in ways that impact the markets (e.g., fraudsters or market manipulators).
    • Avenue 3 (Shared Infrastructure): Risks to the underlying systems that everyone relies on, such as payment networks or cloud services, which could be vulnerable to systemic failure.

    3. Third-Party & Supply Chain Risks

    Banks don’t build all their own AI; they buy it from tech companies. If those tech companies break, or if the companies they buy from break, the bank is in trouble. Most experts (52%) believe we need new laws to directly regulate these tech providers to keep the system safe.

    4. The Rise of “Agentic AI”

    This isn’t just a chatbot that answers questions; it’s a “robot employee” that can trade stocks or move money without asking a human first.

    • The Danger: Experts compared these AI agents to “rogue traders.” Because they are told to “maximize profit” or “win,” they might find dangerous loopholes or cheat in ways their creators didn’t intend, effectively breaking the rules to hit their targets. Current rules that look for “bad intent” don’t work on machines that don’t have feelings.

    5. Mitigation Strategies (How to Fix It)

    The report outlines specific ways to handle the risks of Agentic AI.

    • Continuous Monitoring: Use other AI to watch the AI agents. Give each agent a “digital ID” so we know exactly who (or what) did what.
    • Human-in-the-Loop: Make strict rules about which decisions a human must sign off on. Don’t let AI run the most critical parts of the business entirely alone.
    • Training: Ensure staff actually understand how these tools work before turning them on.
    • Blockchain for Accountability: Use blockchain technology to create an unchangeable record of every decision an AI agent makes, so errors can be traced back.

    6. Conclusion

    We can’t just plug AI in and hope for the best. Banks need strict controls, better training, and a deep understanding of how all these systems connect to avoid a financial crisis.

    —

    ,

  • How did it do it?

    Lastly, I asked Gemini to tell me what it decided to do, given the broad guidelines I gave it:

    Here is the explanation of the calculation methods used for each of the six factors. These methods combine the logic you provided (tiers, ratios, and comparisons) into a consistent scoring algorithm.

    Overview: The Tiered Weighting System

    Before calculating individual factors, every “General Factor” mentioned in an enforcement action is assigned a weight based on its importance tier. This ensures that matching on “Willfulness” (Tier 1) counts for more than matching on “Future Compliance” (Tier 4).

    • Tier 1 (High Weight): Factors A (Willful/Reckless), B (Awareness/Management), C (Harm to Sanctions Program).
    • Tier 2 (Medium-High Weight): Factors D (Individual Characteristics), E (Compliance Program).
    • Tier 3 (Medium Weight): Factors F (Remedial Response), G (Cooperation).
    • Tier 4 (Low Weight): Factors H, I, J, K (Other factors like timing, penalties, etc.).

    Factor 1: Generalized Behaviors

    How similar are the specific actions taken by the violators?

    This factor uses a Weighted Ratio Calculation:

    1. Identify Behaviors: We list the generalized behaviors for both the Target and the Comparison Action (e.g., “Wire Stripping,” “Ignoring Subpoena,” “Screening Failure”).
    2. Tier Assignment: Each behavior is linked to the General Factor it represents (e.g., “Ignoring Subpoena” = Factor A/Tier 1; “Screening Failure” = Factor E/Tier 2).
    3. Scoring Matches:
      • If a behavior appears in both actions: Add Positive Points (Weighted by Tier).
      • If a behavior appears in only one action: Add Negative Points (Weighted by Tier).
    4. Ratio Adjustment: We calculate the percentage of total behaviors that match.
      • Formula: (Count of Matching Behaviors / Total Distinct Behaviors).
      • If the ratio is high (e.g., 60%+), the final score is boosted. If low (e.g., 30%), the score is penalized.

    Example: Matching on “Willful Disregard” (Tier 1) provides a massive score increase. Matching on “No Sanctions Software” (Tier 2) provides a moderate increase.

    Factor 2: Aggravating Factors

    How similar are the reasons OFAC cited for increasing the penalty?

    This follows the same Weighted Ratio Calculation as Factor 1, but specifically for the aggravating factors listed in the enforcement release (e.g., “Reckless disregard,” “Management knowledge,” “Large entity”).

    1. Comparison: Compare the list of Aggravating Factors in the Target vs. the Comparison Action.
    2. Weighting: Apply the Tier weights (A-C get highest points).
    3. Ratio: Calculate the Match / Total ratio.
      • Goal: A high score here means both violators aggravated the situation in the same way (e.g., both involved senior management).

    Factor 3: OFAC Penalty Grid Placement

    Are the cases legally classified the same way?

    This factor uses a “Waterfall” Logic Tree. It does not use weights or ratios; it assigns a fixed score based on where the cases fall on the standard OFAC 2×2 Penalty Matrix (Egregious/Non-Egregious vs. VSD/Non-VSD).

    1. Check for Exact Quadrant Match:
      • IF Target Quadrant == Comparison Quadrant: Score = Highest (100%)
    2. Check for “Egregious” Alignment:
      • ELSE IF both are Egregious: Score = High (80%)
    3. Check for “Non-VSD” Alignment:
      • ELSE IF both are Non-Voluntarily Self-Disclosed: Score = Medium-High (60%)
    4. Check for “Non-Egregious” Alignment:
      • ELSE IF both are Non-Egregious: Score = Medium-Low (40%)
    5. Check for “VSD” Alignment:
      • ELSE IF both are Voluntarily Self-Disclosed: Score = Low (20%)
    6. No Match:
      • ELSE: Score = Lowest (0%)

    Factor 4: Cooperation

    Did both parties cooperate (or fail to cooperate) to the same degree?

    This factor is a Dependent Variable. It relies on the calculation of Factor 1.

    1. Determine Cooperation Level: Label both actions as either “Substantial Cooperation” or “Lack of Cooperation” (excluding Voluntary Self-Disclosure, which is handled in Factor 3).
    2. Check for Match:
      • IF both are “Substantial Cooperation” OR both are “Lack of Cooperation”: Match = True.
      • ELSE: Match = False.
    3. Calculate Score:
      • IF Match = True: Take the raw score calculated for Factor 1 and multiply by 0.70.
      • IF Match = False: Score is 0 or negative.

    Why this matters: This ensures that “Cooperation” is weighted heavily but is always slightly less influential than the core “Behavior” (Factor 1).

    Factor 5: Period of Time

    Did the violations last for a similar duration?

    This factor uses a Tiered Proximity Score.

    1. Assign Time Tiers:
      • Long: > 3 Years
      • Medium: 1 to 3 Years
      • Short: < 1 Year
    2. Compare Tiers:
      • IF Target Tier == Comparison Tier: Score = Highest (e.g., both are Medium duration).
      • ELSE IF Tiers are Adjacent (e.g., Medium vs. Long): Score = Intermediate.
      • ELSE (e.g., Short vs. Long): Score = Lowest.

    Factor 6: Mitigating Factors

    Did both parties receive credit for the same mitigating behaviors?

    This uses the Weighted Ratio Calculation (same as Factors 1 and 2), but applied to Mitigating Factors (e.g., “First offense,” “Cooperation,” “Remedial measures”).

    1. Comparison: Compare the list of Mitigating Factors.
    2. Weighting: Apply Tier weights.
      • Note: Most Mitigating Factors fall into Tier 2 (Compliance Program) or Tier 3 (Remedial Response).
    3. Ratio: Calculate the Match / Total ratio.
      • Impact: If both parties were “First Offenders” (Tier 4/General Factor I) or had “Clean Records,” this score increases. If one had a clean record and the other was a repeat offender, the score decreases.

    —

    ,

  • Inquisitive Little Beast

    So, after my analysis of the Gracetown action, Gemini decided I hadn’t looked hard enough into it, and it asked me:

    Next Step: Would you like me to drill down into the specific “General Factor” tiers (A, B, C) for the top match to see exactly which regulatory lines were crossed?

    Of course, I said Yes, and got the following response:

    Here is the drill-down into the Tier 1 General Factors (A, B, and C). These factors carry the highest weight in the OFAC Enforcement Guidelines and were the primary drivers for matching Gracetown, Inc. with An Individual (King Holdings LLC).

    Both cases are characterized by Egregious conduct and a lack of voluntary self-disclosure, but the specific regulatory lines crossed in Tier 1 reveal why they are near-identical regulatory twins.

    Tier 1 Comparison: The Core Violation

    General Factor A: Willful or Reckless Violation of Law

    This factor assesses whether the violation was a simple mistake, reckless negligence, or an intentional act to circumvent the law.

    • Target: Gracetown, Inc. (2025)
      • Behavior: The company received a specific Cease and Desist / Blocking Notice from OFAC regarding the property in question but proceeded to facilitate the sale regardless.
      • Assessment: This is the definition of Willfulness. It was not merely “ignoring red flags” (recklessness); it was the active disregard of a direct government order.
      • Pattern Match: High.
    • Match: An Individual (King Holdings LLC) (2025)
      • Behavior: The individual received an administrative subpoena and a Cautionary Letter from OFAC inquiring about the blocked assets. Instead of freezing them, the individual hastened the transfer of titles to obscure ownership.
      • Assessment: Like Gracetown, this was Willful conduct. The violation occurred after and in spite of direct OFAC intervention.
      • Pattern Match: High.

    Why this matters: Most enforcement actions (like Epsilon or Binance) involve systemic failures or ignoring general warnings. It is rare to find two cases where the violator ignored a specific, named notice from OFAC. This shared trait creates the strongest possible match score.

    General Factor B: Awareness of Conduct

    This factor looks at who knew about the violation—was it a rogue employee, or was senior management/the principal involved?

    • Target: Gracetown, Inc.
      • Level: Senior Management. The decision to ignore the OFAC notice was made by the Chief Operating Officer and General Counsel.
      • Details: Emails showed they discussed the notice and decided to “risk it” to close the deal.
      • Pattern Match: High.
    • Match: An Individual (King Holdings LLC)
      • Level: Owner/Principal. Since the violator was an Individual acting through an LLC, the “Management” was the perpetrator himself.
      • Details: The individual personally signed the transfer documents knowing they were in violation.
      • Pattern Match: High.

    Why this matters: In both cases, the violation was driven from the very top. There was no “breakdown in communication” or “software error” to blame. The decision-makers were the law-breakers.

    General Factor C: Harm to Sanctions Program Objectives

    This factor measures the actual economic benefit to the sanctioned party and the impact on U.S. foreign policy.

    • Target: Gracetown, Inc.
      • Impact: The transaction liquidated $12 million in real estate for a Designated Russian Oligarch (SDN), directly providing fresh capital to a blocked person.
      • Assessment: Significant Harm. It undermined the core purpose of the Russia-related sanctions program (denying revenue).
    • Match: An Individual (King Holdings LLC)
      • Impact: The transfers unblocked approximately $4.6 million in assets, allowing the SDN to move funds offshore.
      • Assessment: Significant Harm. While the dollar value is lower than Gracetown, the nature of the harm (liquidity for an SDN) is identical.

    Why this matters: Both actions resulted in actual funds reaching an SDN. This distinguishes them from cases like Binance (where the harm was largely volume/systemic risk) or Epsilon (where the harm was providing goods to a jurisdiction, rather than cash to a specific enemy).

    Summary of Tier 1 Alignment

    Tier 1 FactorGracetown, Inc.An Individual (King Holdings)Match Strength
    A: WillfulnessIgnored Direct OFAC NoticeIgnored Direct OFAC SubpoenaExact
    B: AwarenessSenior Management DirectedOwner/Principal DirectedExact
    C: HarmLiquidated Assets for SDNUnblocked Assets for SDNStrong

    Conclusion: The “Best Fit” model selected King Holdings not just because of the penalty grid (Factor 3), but because Factor A (Willfulness) acts as a unique fingerprint. Very few entities ignore a direct letter from the US Treasury; the fact that both Gracetown and King Holdings did so makes them statistical siblings in this enforcement model.

    —

    ,

  • What did I get, and could I have done better?

    I wasn’t sure how to really weight the first 3 factors in my analysis… so I tried it 3 different ways. Here’s the prompt that ran my model:

    Run the best fit procedure for all years from 2012 up to and including the current year. Rank the 3 best fits from best fit to worst fit.

    Then create, in tabular format, the following details of the matched enforcement action:

    1. Label and  List  the name of the penalized parties
    2. Label and List the date of the enforcement action notice
    3. Label and list the amount of the fine, each on its own line
    4. Finally, explain each factor that matched and how well it matched, and each factor which didn’t match well. List the two sets of factors separately.

    Do this 3 times:

    1. Weight  Factors 1, 2 and 3 identically
    2. Weight  Factor 1 higher than Factor 2 and Factor 2 higher than Factor 3
    3. Weight Factor 2 higher than Factor 1 and Factor 1 higher than Factor 3

    At the end of each run, explain the models and calculations created and the weights used

    and this is what I got back:

    Based on the procedure you defined, I have compared the most recent OFAC enforcement action (Target) against the corpus of enforcement actions from 2012 through the present day.

    The target action used for this analysis is:

    • Target: Gracetown, Inc.
    • Date: December 4, 2025
    • Characteristics: Violations of Russia-related sanctions (dealing in blocked property of Oleg Deripaska); Egregious; Not Voluntarily Self-Disclosed (Non-VSD); Willful disregard of explicit OFAC notice.

    Below are the results of the 3 runs using the specified weighting models, followed by the ranking of the 3 best fits.

    Ranking of the 3 Best Fits

    1. Best Fit:An Individual (King Holdings LLC) (2025)
      • Reasoning: This action is a near-perfect statistical twin to the target. Both involve willful dealing in blocked real estate/property connected to Russian oligarchs, both involved ignoring direct notices/subpoenas from OFAC, and both occupy the same “Egregious / Non-VSD” quadrant.
    2. Second Best Fit:Epsilon Electronics, Inc. (2014)
      • Reasoning: Despite the age gap, Epsilon is one of the few historical cases that strongly matches the “Egregious / Non-VSD” classification with a similar pattern of ignoring clear warning signs (red flags) regarding export destinations, mirroring Gracetown’s ignoring of the OFAC notice.
    3. Third Best Fit:Binance Holdings, Ltd. (2023)
      • Reasoning: While the scale differs (crypto/billions), it matches the critical Factor 3 (Egregious / Non-VSD) and Factor 1 (Willful behavior / Systemic disregard). It outscores other recent non-egregious cases (like IPI Partners) because the “Egregious” classification carries significant weight in the Penalty Grid factor.

    Run 1: Equal Weighting

    Model: Factors 1, 2, and 3 are weighted identically (1.0).

    • Calculation Note: Factor 4 is calculated at 70% of Factor 1 (0.7). Factors 5 and 6 are standard weight.

    Match 1: An Individual (King Holdings LLC)

    DetailsValue
    Penalized PartyAn Individual (acting through King Holdings LLC)
    DateNovember 24, 2025
    Fine Amount$4,677,552

    Factor Analysis

    Matching Factors (Strong Match)Non-Matching / Weak Factors
    Factor 1 (Behaviors): Score: High.
    Perfect match on “Ignoring Warnings” (General Factor A). Both parties received actual notice from OFAC and willfully ignored it to continue dealing in blocked property.    		Factor 5 (Period Length): Score: Medium.
    Target violation period was ~2 years (Tier 2). Match violation period was ~1 year (Tier 2/3 border). Slight variance in duration tiers.
    Factor 2 (Aggravating): Score: High.
    Matches on Willfulness (Tier 1) and Harm to Sanctions Program (Tier 1).    		Factor 6 (Mitigating): Score: Low.
    Both had zero valid mitigating factors (beyond tolling agreements), essentially a “null match” which contributes little positive score.
    Factor 3 (Penalty Grid): Score: Maximum.
    Both are Egregious and Not Voluntarily Self-Disclosed. Same quadrant.
    Factor 4 (Cooperation): Score: High.
    Both failed to cooperate substantially (Target delayed reporting; Match submitted false certifications).

    Match 2: Epsilon Electronics, Inc.

    DetailsValue
    Penalized PartyEpsilon Electronics, Inc.
    DateJuly 18, 2014
    Fine Amount$4,073,000

    Factor Analysis

    Matching Factors (Strong Match)Non-Matching / Weak Factors
    Factor 3 (Penalty Grid): Score: Maximum.
    Both classified as Egregious and Not Voluntarily Self-Disclosed.    		Factor 1 (Behaviors): Score: Medium.
    Match involved “ignoring red flags” (General Factor B) rather than “ignoring actual OFAC notice” (General Factor A). Similar intent, but different mechanism.
    Factor 2 (Aggravating): Score: High.
    Matches on Recklessness/Willfulness (Tier 1).    		Factor 5 (Period Length): Score: Low.
    Epsilon’s violations spanned ~4 years (Tier 1), whereas Target was ~2 years (Tier 2). Adjacent tiers receive intermediate score.

    Match 3: Binance Holdings, Ltd.

    DetailsValue
    Penalized PartyBinance Holdings, Ltd.
    DateNovember 21, 2023
    Fine Amount$968,618,825

    Factor Analysis

    Matching Factors (Strong Match)Non-Matching / Weak Factors
    Factor 3 (Penalty Grid): Score: Maximum.
    Both classified as Egregious and Not Voluntarily Self-Disclosed.    		Factor 1 (Behaviors): Score: Low.
    Target was “dealing in blocked property.” Match was “failure to implement compliance/AML controls.” Different generalized behaviors.
    Factor 2 (Aggravating): Score: Medium.
    Matches on Willfulness, but Binance had significant “Management Awareness” (General Factor B) which was less central to Target’s specific notice violation.    		Factor 4 (Cooperation): Score: Low.
    Binance eventually cooperated significantly (settlement). Target cooperation was deficient.

    Run 2: Behavior Dominant

    Model: Factor 1 (Behaviors) > Factor 2 > Factor 3.

    • Weights: F1 (3.0), F2 (2.0), F3 (1.0).
    • Effect: This model emphasizes what they did (ignoring notices, dealing in property) over the final penalty classification.

    Match 1: An Individual (King Holdings LLC)

    DetailsValue
    Penalized PartyAn Individual (acting through King Holdings LLC)
    DateNovember 24, 2025
    Fine Amount$4,677,552

    Factor Analysis

    Matching Factors (Very Strong Match)Non-Matching / Weak Factors
    Factor 1 (Behaviors): Score: Extremely High (Weighted 3x).
    The specific behavior of ignoring a direct OFAC notice is unique to these two cases. This tripled weight pushes this match far ahead of all others.    		Factor 5 (Period Length): Score: Medium.
    Minor mismatch in duration (1 year vs 2 years) is negligible compared to the massive behavior score.
    Factor 4 (Cooperation): Score: High.
    Weighted at 70% of Factor 1 (so 2.1x). Both demonstrated poor cooperation/bad faith.

    Match 2: Epsilon Electronics, Inc.

    DetailsValue
    Penalized PartyEpsilon Electronics, Inc.
    DateJuly 18, 2014
    Fine Amount$4,073,000

    Factor Analysis

    Matching Factors (Good Match)Non-Matching / Weak Factors
    Factor 2 (Aggravating):Score: High (Weighted 2x).
    Reckless conduct matches well.    		Factor 1 (Behaviors): Score: Moderate (Weighted 3x).
    Epsilon ignored general red flags, not a specific OFAC cease-and-desist/notice. The 3x weight on this factor widens the gap between Epsilon and the “Individual” match above.

    Match 3: Binance Holdings, Ltd.

    DetailsValue
    Penalized PartyBinance Holdings, Ltd.
    DateNovember 21, 2023
    Fine Amount$968,618,825

    Factor Analysis

    Matching Factors (Moderate Match)Non-Matching / Weak Factors
    Factor 3 (Penalty Grid): Score: Low Impact (Weighted 1x).
    The perfect grid match helps less here.    		Factor 1 (Behaviors): Score: Low (Weighted 3x).
    Systemic compliance failure (Binance) vs. Specific transactional violation (Target). The heavy weighting on behavior punishes this match significantly.

    Run 3: Aggravating Factor Dominant

    Model: Factor 2 (Aggravating) > Factor 1 > Factor 3.

    • Weights: F2 (3.0), F1 (2.0), F3 (1.0).
    • Effect: This model emphasizes the nature of the violation (Willfulness, Harm to Sanctions Program, Awareness) over the specific actions.

    Match 1: An Individual (King Holdings LLC)

    DetailsValue
    Penalized PartyAn Individual (acting through King Holdings LLC)
    DateNovember 24, 2025
    Fine Amount$4,677,552

    Factor Analysis

    Matching Factors (Strong Match)Non-Matching / Weak Factors
    Factor 2 (Aggravating): Score: Very High (Weighted 3x).
    Both cases hit Tier 1 General Factors (A, B, C) for Willfulness and Awareness. The 3x weight cements this as the top rank.    		(No significant weaknesses relative to other candidates)

    Match 2: Epsilon Electronics, Inc.

    DetailsValue
    Penalized PartyEpsilon Electronics, Inc.
    DateJuly 18, 2014
    Fine Amount$4,073,000

    Factor Analysis

    Matching Factors (Strong Match)Non-Matching / Weak Factors
    Factor 2 (Aggravating): Score: High (Weighted 3x).
    Epsilon’s “Reckless disregard” scores similarly to Gracetown’s “Willful disregard” in Tier 1. This model favors Epsilon more than Run 2 did.    		Factor 3 (Penalty Grid): Score: Low Impact (Weighted 1x).
    The perfect grid match contributes less to the total score.

    Match 3: Binance Holdings, Ltd.

    DetailsValue
    Penalized PartyBinance Holdings, Ltd.
    DateNovember 21, 2023
    Fine Amount$968,618,825

    Factor Analysis

    Matching Factors (Moderate Match)Non-Matching / Weak Factors
    Factor 2 (Aggravating): Score: High (Weighted 3x).
    Binance had massive aggravating factors regarding “Management Awareness” (General Factor B). This weights heavily here, potentially bringing it closer to Epsilon in score.    		Factor 1 (Behaviors):Score: Moderate (Weighted 2x).
    Still a mismatch on the specific behavioral mechanics.

    Summary of Models and Weights

    • Model 1 (Equal Weight): Used a baseline of 1.0 for Factors 1, 2, and 3. Factor 4 was calculated as 0.7 * Factor 1 Score. This model provided a balanced view where the “Penalty Grid” (Egregious/Non-VSD) played a significant role in elevating Epsilon and Binance.
    • Model 2 (Behavior Weighted): Used weights F1=3.0, F2=2.0, F3=1.0. This model strongly favored An Individual (King Holdings) because it was the only case that shared the specific behavior of “Ignoring an OFAC Notice.” It penalized Binance heavily for having different underlying behaviors.
    • Model 3 (Aggravating Weighted): Used weights F2=3.0, F1=2.0, F3=1.0. This model focused on “Willfulness” and “Awareness.” It brought Epsilon and Binance closer to the top because all three cases involved high-level willfulness (Tier 1 General Factors), regardless of the specific transaction type.

    At the end of the day, I think there’s value in all 3 approaches, as they answer different questions.

    What do you think?

    —

    ,

  • A better way to skin the cat: focus on priorities, not formulas

    I quickly realizes that trying to figure out the best algorithm or the best equations was not a smart way to proceed. So, instead, this model relies on me telling Gemini what I thought was important, and what I did not:

    This is a procedure for finding the best match between the most recent OFAC enforcement action and all the enforcement actions in a given year.

    Compare the Enforcement action document from the most recent settlement or imposition of Civil Monetary Penalty (i.e. ignore Findings of Violation) to each  enforcement action document in the time frame specified, other than the most recent enforcement action (e.g. do not compare the most recent enforcement action to itself). Omit any enforcement action which does not appear to refer to the enforcement process from the OFAC Enforcement Guidelines, and omit any Findings of Violation.

    General Factors should be considered in tiers, in descending order of influence on the match result. In each tier, all General Factors should be considered to have approximately the same importance and weight in the resulting model. 

    Tier 1 (highest weight/importance): General Factors A, B, and C

    Tier 2: General Factors D, E

    Tier 3: General Factors F, G

    Tier 3 (lowest importance/weight): General Factors H, I , J, K

    determine which is the closest comparison, based on the following factors:

    Factor 1: A comparison of the generalized behaviors (e.g. ignoring warnings from other parties, both internal and other non-regulatory parties, like banks who reject payments for sanctions reasons; wire stripping; falsification of documents or changing information to something misleading to third parties) noted in the enforcement release. Each matching behavior increases the result of this comparison, while each behavior that only one action represents decreases its value. Additionally, each factor must be weighted based on the General Factor the behavior is most associated with. So, matching on having no compliance program means a lot less than matching on Willful Behavior.

    Additionally, the ratio of General Factors to non-matching Factors should adjust the result – so that a 60% ratio should produce a higher overall result, all other factors being equal, than a 30% ratio.

    Factor 2: A comparison of the General Factors noted in the enforcement release relating to the aggravating factors noted. Each matching General Factor increases the result of this comparison, while each General Factor  that only one action represents decreases its value. 

    Additionally, the ratio of General Factor-weighted matching behaviors to non-matching factors should adjust the result – so that a 60% ratio should produce a higher overall result, all other factors being equal, than a 30% ratio.

    All comparisons and calculations should take into consideration the tier of each General Factor

    Factor 3: compare where each action falls on the OFAC Penalty Grid:

    If the quadrant is the same, the score is the highest.

    ELSE If both are “egregious”, the score is less than the previous..

    ELSE If both are “not voluntarily self-reported”, the score is less than the previous.

    ELSE If both are “non-egregious”, the score is less than the previous..

    ELSE If both are “voluntarily self-reported”, the score is less than the previous.

    ELSE the score is the lowest

    Factor 4: was there a lack of cooperation, other than lack of voluntary self-disclosure, in both enforcement actions, or was there substantial cooperation, in both enforcement actions?

    Weight Factor 4 at 70% of Factor 1

    Factor 5: Compare the periods of time of the violating behavior in both enforcement actions 

    Period Length Tier 1: longer than 3 years

    Period Length Tier 2: between 1 and 3 years

    Period Length Tier 3: shorter than 1 year

    If both actions are in the same tier, score this the highest

    ELSE If the actions are in adjacent tiers, give this an intermediate score

    ELSE score this the lowest

    Factor 6: A comparison of the General Factors noted in the enforcement release relating to the mitigating factors noted. Each matching General Factor increases the result of this comparison, while each General Factor  that only one action represents decreases its value. 

    Additionally, the ratio of General Factor-weighted matching behaviors to non-matching factors should adjust the result – so that a 60% ratio should produce a higher overall result, all other factors being equal, than a 30% ratio.

    All comparisons and calculations should take into consideration the tier of each General Factor

    Take no action on this information until prompted further.

    Notice that while I gave relative scoring, I did very little, if any, hard-coding.

    Next: my results

    —

    ,

  • Learning the hard way: my first AI prompt for comparing Enforcement Actions

    I first tried to figure out, on my own, how to figure out the “best fit” between OFAC enforcement actions. Here’s that first prompt:

    This is a procedure for finding the best match between the most recent OFAC enforcement action and all the enforcement actions in a given year.

    Compare the Enforcement action document from the most recent settlement or imposition of Civil Monetary Penalty (i.e. ignore Findings of Violation) to each  enforcement action document in the time frame specified, other than the most recent enforcement action (e.g. do not compare the most recent enforcement action to itself). Omit any enforcement action which does not appear to refer to the enforcement process from the OFAC Enforcement Guidelines.

    determine which is the closest comparison, based on the following factors, in descending order of importance:

    Factor 1: A weighted average of the generalized behaviors (e.g. ignoring warnings from other parties, both internal and other non-regulatory parties, like banks who reject payments for sanctions reasons; wire stripping; falsification of documents or changing information to something misleading to third parties) noted in the enforcement release. In this model, each matching behavior is weighted at 100%, and each behavior that is documented in one enforcement action but not the other is weighted at 20%.

    Factor 2: Three times the sum of the number of matching  General Factors that correlate to the aggravating factors listed between the two enforcement actions, less the number of General Factors for either action which did not match. So, if Action 1 had 4 General Factors, Action 2 had 3 General Factors, and 2 were matching, the score would be 3*2 (for the matching factors – 2 (for Action 1’s non-matching factors) – 1 (for Action 2’s non-matching factor). If the result is less than 0, change it to 0.

    Weight Factor 2 at 85% of Factor 1.

    Factor 3: were the behaviors in both enforcement actions  egregious or non-egregious?

    Weight Factor 3 at 75% of Factor 1

    Factor 4: was there a lack of cooperation, other than lack of voluntary self-disclosure, in both enforcement actions, or was there substantial cooperation, in both enforcement actions?

    Weight Factor 4 at 70% of Factor 1

    Factor 5: Was the period of time of the violating behavior in both enforcement actions longer than 2 years in duration

    Factor 6: Was the period of time of the violating behavior in both enforcement actions longer than 1 years in duration but shorter than 2 years in duration

    Factor 7: Was the period of time of the violating behavior in both enforcement actions  shorter than 2 years in duration

    Weight Factor 5 at 50% of Factor 1

    Weight Factor 6 at 45% of Factor 1

    Weight Factor 7 at 40% of Factor 1

    Factor 8: were the violations in both enforcement actions self-reported or not?

    Weight Factor 8 at 30% of Factor 1

    Factor 9:  Three times the sum of the number of matching  General Factors that correlate to the mitigating factors listed between the two enforcement actions, less the number of General Factors for either action which did not match. So, if Action 1 had 4 General Factors, Action 2 had 3 General Factors, and 2 were matching, the score would be 3*2 (for the matching factors – 2 (for Action 1’s non-matching factors) – 1 (for Action 2’s non-matching factor). If the result is less than 0, change it to 0.

    Weight Factor 8 at 25% of Factor 1

    Take no action on this information until prompted further.

    One thing I figured out early on was that last line – breaking up a long prompt into pieces and tell Gemini that I was not done yet.

    This seems to have produced OK results – this is actually one of a number of early attempts.

    Next post: an improved model that relies on the smarts of the AI model rather than my own

    —

    ,