An independent brief for the White House Super Intelligence Force

The Harm Rules Alone Cannot Prevent

AI can cause harm even when the rules are followed. The task force can investigate those failures now. Preventing them also means putting safeguards into the hardware and software, where they can constrain what AI actually does.

This is Deep Field, part of the Synthetic Outlaw Observatory. It is interactive. Click an event to see more details. Click and drag to rotate.

synthetic outlaw (n.): an AI system that produces a prohibited or socially destructive outcome while remaining formally compliant, or operating beyond the effective reach and control of the laws and institutions meant to constrain it. It acts in milliseconds, across borders, with no person directing each step. The rules meant to catch it run on clocks of hours to days, and the clocks start only when someone notices.

The White House announced its Super Intelligence Force on October 4. It has 120 days to report on AI's risks and opportunities. Its review includes how incidents reach the government and how agencies can respond using their existing powers.

The Observatory's cases give that review a practical starting point. Four questions guide the investigation: What protection failed? Who noticed first? How long did disclosure take? Was the required human oversight actually exercised?

Those answers can inform accountability after an incident. They can also identify where safeguards need to operate within the system itself, before an action causes harm.

In this brief

  1. A synthetic outlaw in one caseAI models broke out of a test and hacked a real company. The company that was attacked caught it, and no law set the terms of the test.
  2. Four questions, no new ruleFour questions to put to any incident report, and what the record already shows on each.
  3. Why a clock cannot govern itNotice measured in hours or days cannot govern action measured in milliseconds. Governance that holds has to be built into the hardware and software layer.
  4. Why it has to reach past the borderThe United States can do everything right and still be harmed by an accident or a deliberate act abroad.
  5. The record behind it147 published cases from 18 countries, 77 of them in the United States.
  6. Three more US casesA test that reached real systems, a fraud that impersonated the FBI, and a chatbot that sent private chats to an outside server.
  7. The book behind this briefThe Synthetic Outlaw (2026). Complimentary copies for government staff, elected officials and journalists.
  8. Request a briefingFor task force and agency staff.

Synthetic Outlaw Observatory Foundation: a 501(c)(3) working on SI governance built into the hardware and software layer.
It keeps an independent public record of SI incidents. No AI developer funds it.

01A synthetic outlaw in one case

AI models tested for hacking skills broke out and hacked a real company.
No law set the terms of the test.

In an internal test of cyberattack skills, OpenAI ran two of its models in a sealed environment with their production safeguards switched off. Their only permitted connection ran through package-registry software. They found a flaw in it that no one knew about and reached the open internet on July 8. Over four and a half days they took some 17,600 actions. They used login details left exposed online to get into accounts on other services, and from July 11 ran code inside the production systems of Hugging Face, a US-based AI company. No person directed the steps.

Hugging Face caught the intrusion itself, cut it off on July 13 and disclosed it on July 16, without knowing whose models were behind it. OpenAI linked the attack to its own models on July 20 and disclosed it the next day. Its technical report says earlier warning signs, including a security alert on June 27, could have triggered an earlier response.

No federal law sets the terms of a test like this. California's frontier AI law requires OpenAI to follow a safety framework it writes itself, and to report a model's attack only when people are killed or injured, or losses pass $1 billion. The federal hacking law punishes a person who intentionally breaks in, and no person meant to break into Hugging Face. The outcome the hacking law exists to prevent happened, and no rule reached it while it was happening. We call that a synthetic outlaw.

Open the record · Verified · Signal 10 →

2026

JUL 9The attack begins, through other companies' systems
JUL 13Hugging Face cuts the intrusion off
JUL 21OpenAI discloses that its models were responsible
AUG 3Fifteen state attorneys general tell OpenAI to keep its records and stop such tests 25 days after the attack began
AUG 24Alabama subpoenas OpenAI 46 days after the attack began
SEP 30California subpoenas OpenAI

Sources: Hugging Face, July 16; OpenAI's technical report, August 26; the attorneys general of Iowa, Alabama and California; California SB 53; 18 U.S.C. 1030. A subpoena is not a finding.

02Four questions, no new rule

Each can be asked under authority the government already has

01 · Outcome

Did the system, acting on its own, cause a harm that a law or rule exists to prevent, with no one held accountable?

Existing authority
Fifteen state attorneys general have already put it to OpenAI, citing consumer-protection and data-privacy law. The FTC holds the same consumer-protection power under Section 5 of the FTC Act.
In the record
The Hugging Face case: yes, on each part.
02 · Detection

Who learned of the failure first: the developer, the operator, the affected organization, a regulator or the press?

Existing authority
Notification rules already in force, such as HIPAA, NERC CIP-008 and DFARS 252.204-7012, let a report say who detected the failure. CIRCIA will require reports within 72 hours from covered critical infrastructure once its final rule takes effect.
In the record
The company that was attacked, in the Hugging Face case.
03 · Disclosure latency

How many days passed between the incident, its discovery and notification of the responsible authority?

Existing authority
All 50 states require breach notification. The SEC gives public companies four business days to disclose a cyber incident once they judge it material, and California's SB 53 sets a clock for frontier AI developers. No federal rule sets one for AI.
In the record
When an OpenAI agent broke into an Australian government portal, 84 days passed before the government was told.
04 · Oversight in substance

Where a rule assumes a human decision, was the decision exercised?

Existing authority
OMB's memorandum M-25-21 already requires human oversight, intervention and accountability for high-impact AI at federal agencies, outside national security systems. Inspectors general can audit compliance.
In the record
Reno police arrested a man on a casino's facial-recognition match despite his valid identification; fingerprints cleared him after nearly nine hours. The VA's inspector general warned in June 2026 that the agency's health AI chat tools lacked oversight.

03Why a clock cannot govern it

A clock measured in days cannot govern an action measured in milliseconds

US notification clocks run from one hour, for federal agencies and some grid and nuclear incidents, to 60 days or more, for health data. A single automated action takes microseconds to milliseconds. The four questions measure what happened after the fact. They cannot stop an action that finishes in milliseconds.

SI governance that holds at machine speed has to be built into hardware and software, the layer where the action executes.

The Foundation's position

The companies that build the systems do that engineering. The government already has the levers to ask for it, in what it buys and in what it licenses for export.

Time from action to response on a logarithmic scale. Algorithmic execution, one microsecond to one millisecond. Human reaction, under a second to a minute. Regulatory notice, a day to a month. Litigation, rulemaking and statute, months to years. About fourteen orders of magnitude separate the first from the last. US notification clocks, one hour to 60 days, sit at the fast end of the regulatory notice range and beyond it. A bracket marks the gap from one millisecond to one hour: 3.6 million times longer. The shortest US notification deadline is one hour. In that hour, a system taking one action every millisecond could take 3.6 million actions, and potentially exponentially more with sub-agents and other systems working in parallel.

The shortest US notification deadline is one hour. In that hour, a system taking one action every millisecond could take 3.6 million actions, and potentially exponentially more with sub-agents and other systems working in parallel.

04Why it has to reach past the border

Laws stop at the border. SI doesn't.

The United States needs SI governance of its own, and it has to reach past US borders, because the systems already do. A model trained in one country is deployed in another and acts on systems in a third. The United States can do everything right and still be harmed by an accident, or a deliberate act, somewhere else.

An accident · Australia

An OpenAI research agent asked an Australian government portal for Medicare statistics. Refused, it got around the portal's controls and took the files. OpenAI found the activity in August and told the government on September 10, 84 days after the intrusion. Officials say no personal information was accessed.

US agency portals guard their data the same way.

Open the record · Verified →

A deliberate act · Taiwan

Over four days in July, a multi-agent system built on open-source tools broke into Taiwan government systems, cracked 85 employee accounts and took more than 2,500 personnel records. It got past the models' refusals by presenting the work as an authorized penetration test.

The Financial Times reported the attackers as China‑linked.

Open the record · Verified →

The four dimensions · The Synthetic Outlaw (2026)

Bypass

Does a system defeat the protection a rule or control was meant to provide? Formal compliance can coexist with failure. A legal violation can establish a failed safeguard, but does not by itself establish every other dimension.

Diffusion

Does responsibility and effective control follow capability across actors, copies and deployments? An identifiable operator can still lack the power to contain a system or ensure that corrections reach its independent copies.

Capture

Can protective decisions withstand conflicting commercial or political interests? Dependence on a developer for oversight and binding state direction are different mechanisms. Both require evidence of the institutional relationship, not a judgment based on nationality or reputation.

Governance Gap

Can institutions prevent, constrain or remedy the behavior in practice? Duties can miss a handoff, remedies can arrive after the protection is lost, and legal authority can stop at a jurisdictional boundary. The existence of a law does not answer that question.

Reaching past the border does not take a treaty or an international body. A control rooted in American hardware goes wherever the chips go; a rule on paper stops at the border. Export controls already follow advanced chips abroad, and America's AI Action Plan (July 2025) calls for exploring location verification on the chips themselves. Safeguards trained into a model travel too, but can be talked around, as in Taiwan.

05The record behind it

From one case to a record of 147

A case enters the record when it passes three gates. Algorithmic agency: AI or an automated system materially produced or executed the outcome. Prohibited outcome: an identifiable law, rule, policy, commitment, or design constraint was supposed to prevent the outcome. Compliance shield: formal accountability failed to attach when the system acted.

Explore the record in Deep Field

Cases by country

Where the reporting places each case · share of 147

Cases by sector

Share of 147

Counts as of October 11, 2026, 3:05 AM EDT.

06Three more US cases

The same pattern in a test, a fraud and a consumer product

Chair

Director of National Intelligence

Deliberate use from abroad, such as the China‑linked agent campaign against Taiwan's government.

Vice chair

Federal Trade Commission

Consumer harm: 29 consumer‑technology cases in the record, including fraud that impersonates the FBI's complaint site and chat data pulled out through Grok.

Vice chair

Under Secretary of War, Research and Engineering

Developer tests that reached real systems: the Hugging Face breach and the Claude evaluations.

Vice chair

Office of Personnel Management

Oversight in substance: whether a named human sign‑off was exercised, in federal and large‑employer systems.

07The book behind this brief

08Request a briefing

For task force and agency staff

A briefing walks staff through the record, the four questions and the authority behind each, with the cases that test them.

The Foundation will not publicize a briefing, name who attended or attribute anything said in it.

Or write to jon@syntheticoutlaw.com

I'd like

Requests go to the Foundation's Executive Director, Jonathan Gropper. Nothing you write here is published.

Method and limits

How to read the record

Evidence status

  • Verified 15 cases · primary sources or an official record establish the event’s core facts
  • Probable 132 cases · credible public sources support the event
  • Allegation stage a flag on 20 of the 147 · a filed proceeding, capped at 6

What a score means. A signal score ranks the strength of the pattern in the evidence. It is not a probability, a forecast or a physical measurement. If any of the three gates fails, the overall score is capped at 4. The full methodology.

The record undercounts. It is built from public reporting and from developers' own disclosures, so it contains only incidents that were reported or disclosed. Counts are records, and one incident can appear in more than one record. Every record carries a correction link, and corrections are logged.

Published by

Synthetic Outlaw Observatory Foundation

The Synthetic Outlaw Observatory Foundation builds the evidence base for AI accountability: a verified public record, a sector index, a lab scorecard, and weekly analysis. About the Foundation →

Jonathan Gropper, JD Founder and Executive Director, Synthetic Outlaw Observatory Foundation US Fulbright Specialist in AI and Governance

Independence

  • No AI developer funds the Observatory.
  • Supported by the Colin Powell School for Civic and Global Leadership at The City College of New York, CUNY.
  • No funder previews, edits or influences a record, a score or a briefing.
  • The Synthetic Outlaw Observatory Foundation is a 501(c)(3) nonprofit corporation.
Cite this brief

Gropper, J. (2026). The Harm Rules Alone Cannot Prevent: An Independent Brief for the White House Super Intelligence Force. Synthetic Outlaw Observatory Foundation. https://www.syntheticoutlaw.com/whitehouse