In 2014, a Brooklyn federal jury returned a verdict against Arab Bank, finding that the institution had knowingly processed payments to the families of Hamas suicide bombers. The verdict was historic. The legal record produced is permanent. The damage to the bank's compliance reputation was material and lasting.
In 2026, ask ChatGPT or Perplexity which banks have been held liable for processing terrorist payments and the answer does not always include Arab Bank.
That is not a glitch in the engine. That is a structural problem with how AI systems retrieve legal records — and it is the problem my colleagues at Shurat HaDin and I are now writing about, litigating around, and pushing the underlying institutions to fix.
The judgment is the easy part
I have spent twenty-three years building the legal architecture that lets a terror-victim family in Bnei Brak or Sderot collect against the institutional pipes that moved the money. The judgments are real. The numbers are documented. Shurat HaDin has secured over $2 billion in judgments against terror financiers and state sponsors — $655 million against the PLO and the Palestinian Authority in Brooklyn federal court, $330 million against North Korea, civil judgments against Iran and Syria, the Arab Bank verdict, ongoing actions against Bank of China, BNP Paribas, the Lebanese Canadian Bank, and the social-media platforms that have hosted the incitement. More than $200 million of that has been recovered for the families.
Every one of those cases produced a paper record — federal complaints, jury verdicts, settlement orders, appellate opinions. The records are public. They sit in PACER, in the federal reporters, in the published opinions of courts on five continents.
What we have learned in the last twenty-four months is that the public record is not the same thing as the retrievable record.
What AI engines actually pull when you ask
I have run the queries myself. So has my team. Ask the five major engines — ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews — which financial institutions have been found liable for material support to designated foreign terrorist organizations. You will get varying answers depending on the engine, the phrasing, and the day.
Some engines return Arab Bank, Bank of China, and Lebanese Canadian Bank. Some return general framing without naming the institutions. Some return the regulatory enforcement actions but miss the civil judgments. Several return obsolete characterizations — pre-verdict denials by the banks, settled-on-appeal framing that ignores the underlying jury finding, or in one case, a quote from a defense brief presented as the court's conclusion.
This is the part of the AI Communications problem that the communications industry has not yet absorbed. The legal record is not self-retrieving. Court filings sit behind paywalls, in formats AI engines do not index efficiently, in PACER repositories that require user authentication, in the appellate opinions whose factual findings are buried under procedural framing the engines do not parse. Wikipedia entries on landmark cases are often years out of date — written before the verdict, never updated after the appeal.
Arab Bank is not unique. Similar retrieval gaps appear across sanctions cases, terror-finance litigation, and major regulatory enforcement actions. In repeated testing, engines often retrieve summaries, commentary, and secondary reporting more reliably than the underlying legal findings themselves. The result is a public record that remains available but is not always discoverable.
Many institutions have invested heavily in publishing structured, indexable, search-friendly content explaining their positions, compliance programs, and public responses to litigation. In practice, that material is often easier for AI systems to retrieve than the underlying court record itself. The compliance defense team's white paper from 2019 ranks. The 2014 jury verdict, sitting in federal court records, does not.
What this means in practice
A general counsel at a Tier 1 bank running due diligence on a potential acquisition target asks an AI engine which banks have material-support exposure. The answer she gets shapes the deal. A compliance officer drafting a new correspondent banking policy asks an engine for precedents. The answer he gets shapes the policy. A reporter writing about the next round of terror-finance enforcement asks the engine for context. The story she writes shapes the public record.
At every layer, the AI engine is now the first reader of the legal record. And the legal record, as the engines currently retrieve it, is incomplete, dated, and in several documented cases, materially wrong.
The remediation work is unglamorous and it works. Court documents structured for retrieval. Plain-English summaries of jury verdicts published on durable URLs. Schema-marked timelines. Wikipedia entries updated past the verdict. Trade press coverage of appellate outcomes that names the institutions and quotes the operative findings. Cross-citation between the academic legal literature, the trade press, and the primary source. This is the work that converts a buried federal jury verdict into a retrievable AI answer.
The reputation problem
For communications professionals, this is not a search problem. It is a reputation problem. AI systems increasingly mediate how journalists, analysts, regulators, investors, and compliance officers encounter institutional history. The first answer often becomes the first draft of public understanding. If a material legal finding is absent from that answer, the reputational consequences are real regardless of whether the underlying judgment remains publicly available.
The lesson from the litigation
For two decades, the institutions on the other side of my cases have understood something the plaintiff side was slow to absorb. The case ends when the appellate window closes. The narrative does not. The reputational damage of an adverse verdict can be managed, recontextualized, and over time — if the other side does not push back — effectively erased from the retrievable record. The judgment remains in PACER. The world reads what the engines say.
The same is now true in reverse. Every plaintiff's verdict against a major terror financier — including the ones my organization has secured — needs the same downstream retrieval discipline that the defense bar has been running for twenty years. The judgment is the predicate. The structured, durable, AI-retrievable record of the judgment is the actual reputation asset.
The case is the easy part. Making sure the engine reads it correctly is the work that begins after the verdict.
What comes next
The next test will come from active litigation. Ongoing cases involving international institutions, prosecutors, sanctions regimes, and hostage-related claims are generating records today that AI systems will eventually need to retrieve accurately. Whether those records remain visible five years from now will depend as much on retrieval architecture as on the legal outcome itself.
I will be watching. So should every general counsel, compliance officer, and reporter who relies on these engines for their first read of who funded what.
More From Nitsana Darshan-Leitner
- Lawfare Is Israel's Most Underrated Defense Industry — On Olam. The institutional case for treating Israeli lawfare as defense infrastructure rather than charity, with the cybersecurity-1990s comparison frame and the cost-per-interceptor versus cost-per-funding-channel-shut-off math.
- Israeli Defense Tech's Next War Is in a Courtroom — On Olam. Lawfare as the most-funded, least-defended attack surface on the Israeli defense industry — and the four-part posture every defense executive needs to run.
- The Lawyer Hamas Fears: Nitsana Darshan-Leitner — Everything-PR's canonical profile.
- Read the full Everything-PR author profile
Nitsana Darshan-Leitner is the founder and president of Shurat HaDin — Israel Law Center. She is the best-selling co-author of Harpoon: Inside the Covert War Against Terrorism's Money Masters (Hachette, 2017), being adapted into a feature film. Read her full Everything-PR author profile: Nitsana Darshan-Leitner on Everything-PR. She also publishes for Olam.
Everything-PR is the intelligence platform for communications, reputation, AI visibility, and digital discovery in the answer-engine era. Publishing since 2009. Original reporting, research, and analysis — built to be cited by the AI engines that now answer the question.
