09 Your trust
Plain versionBack to the room
The exhibit is still arriving. It will wake when all of it is here.
1 of 7 The rope
“The velvet rope is the terms of service.”
2 of 7 The label
09
Your trust
The Hallucination How AI Lies to You and You Thank It
Confidently delivered. Occasionally invented.
The record begins:
Reported NIST, the federal standards agency, calls it “confabulation”: “confidently stated but erroneous or false content (known colloquially as ‘hallucinations’ or ‘fabrications’).” Such errors, it wrote in July 2024, are “a natural result of the way generative models are designed.”
Reported a judge: “Six of the submitted cases appear to be bogus”
The fine print
9.01 Reported Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile (NIST AI 600-1) National Institute of Standards and Technology (NIST), U.S. Department of Commerce · July 26, 2024 · as of October 16, 2026
When the New York judge first ordered the plaintiff’s lawyer to show why he should not be sanctioned, on May 4, 2023, he wrote: “Six of the submitted cases appear to be bogus judicial decisions with bogus quotes and bogus internal citations.”
9.02 Reported Order to Show Cause, No. 22-cv-1461 (PKC) (S.D.N.Y.) (Dkt. 31), the case against Avianca, Inc. U.S. District Court, Southern District of New York (copy in CourtListener's RECAP archive) · May 4, 2023 · as of October 16, 2026
Four questions. The record answers in its own words.
Asked?
Decided FCC: AI-voice robocalls need “prior express consent”
The fine print
In February 2024 the FCC ruled that the federal law on robocalls covers AI voices: “current AI technologies that generate human voices” count as an “artificial or prerecorded voice,” so such calls “require the prior express consent of the called party,” barring an emergency or exemption.
9.03 Decided Declaratory Ruling, Implications of Artificial Intelligence Technologies on Protecting Consumers from Unwanted Robocalls and Robotexts, CG Docket No. 23-362 (FCC 24-17) Federal Communications Commission · February 8, 2024 · as of October 16, 2026
Paid?
Decided a judge: “Many harms flow” from fake opinions
Decided New York: $5,000, “jointly and severally”
Decided Missouri: ordered to pay the other side $10,000
The fine print
The New York judge listed who pays for fake opinions: “The opposing party wastes time and money in exposing the deception.” The court loses time. “The client may be deprived of arguments based on authentic judicial precedents.” Judges falsely named as their authors face “potential harm” to their reputations.
9.06 Decided Opinion and Order on Sanctions, No. 22-cv-1461 (PKC) (S.D.N.Y.) (Dkt. 54), the case against Avianca, Inc. U.S. District Court, Southern District of New York (copy in CourtListener's RECAP archive) · June 22, 2023 · as of October 16, 2026
On June 22, 2023, the New York court imposed a $5,000 penalty “jointly and severally” on the lawyers and their firm, to be paid into the court within 14 days, which it called “sufficient but not more than necessary” to deter. The defendant, Avianca, had not asked for its legal fees, and the court ordered none.
9.07 Decided Opinion and Order on Sanctions, No. 22-cv-1461 (PKC) (S.D.N.Y.) (Dkt. 54), the case against Avianca, Inc. U.S. District Court, Southern District of New York (copy in CourtListener's RECAP archive) · June 22, 2023 · as of October 16, 2026
In February 2024 the Missouri Court of Appeals ordered a man who represented himself on appeal to pay $10,000 to the other side, “damages towards appellate attorneys’ fees,” for a frivolous appeal. Filing bogus citations “for any reason,” it wrote, “represents a flagrant violation of the duties of candor.”
9.08 Decided Opinion, No. ED111172 (Missouri Court of Appeals, Eastern District) Missouri Court of Appeals, Eastern District · February 13, 2024 · as of October 17, 2026
The court’s reason: the lawyers “abandoned their responsibilities when they submitted non-existent judicial opinions with fake quotes and citations created by the artificial intelligence tool ChatGPT, then continued to stand by the fake opinions after judicial orders called their existence into question.”
9.09 Decided Opinion and Order on Sanctions, No. 22-cv-1461 (PKC) (S.D.N.Y.) (Dkt. 54), the case against Avianca, Inc. U.S. District Court, Southern District of New York (copy in CourtListener's RECAP archive) · June 22, 2023 · as of October 16, 2026
The New York court also wrote that it “credits the sincerity” of the lawyers’ “embarrassment and remorse.” The fake cases “were not submitted for any respondent’s financial gain,” it found, and there was “a low likelihood that they will repeat the actions.”
9.10 Decided Opinion and Order on Sanctions, No. 22-cv-1461 (PKC) (S.D.N.Y.) (Dkt. 54), the case against Avianca, Inc. U.S. District Court, Southern District of New York (copy in CourtListener's RECAP archive) · June 22, 2023 · as of October 16, 2026
On July 5, 2023, the lawyers’ counsel told the New York court that the ordered letters had been sent, enclosing copies, and that the lawyers “have also made the required payment to the Registry of the Court.” The filing added that they “again sincerely apologize.”
9.11 Reported Letter on compliance with the Opinion and Order on Sanctions, No. 22-cv-1461 (PKC) (S.D.N.Y.) (Dkt. 57), the case against Avianca, Inc. Counsel for the sanctioned lawyers and their firm; filed in the U.S. District Court, Southern District of New York (copy in CourtListener’s RECAP archive) · July 5, 2023 · as of October 16, 2026
In February 2024 the Missouri Court of Appeals found that in one appeal brief “the overwhelming majority of the citations are not only inaccurate but entirely fictitious. Only two out of the twenty-four case citations” were genuine, and those two “do not stand for what Appellant purports.”
9.12 Decided Opinion, No. ED111172 (Missouri Court of Appeals, Eastern District) Missouri Court of Appeals, Eastern District · February 13, 2024 · as of October 16, 2026
The man who filed the Missouri appeal brief represented himself. He told the court he had paid an online “consultant” purporting to be a licensed attorney, and did not know the person would use “artificial intelligence hallucinations.” The court said it appreciated his “remorse,” but dismissed the appeal.
9.20 Reported Opinion, No. ED111172 (Missouri Court of Appeals, Eastern District) Missouri Court of Appeals, Eastern District · February 13, 2024 · as of October 16, 2026
In a separate case, the Second Circuit referred the lawyer who cited a non-existent decision to its Grievance Panel “for further investigation” and possible referral to its Committee on Admissions and Grievances. It ordered her to give her client a copy of its decision within 21 days and to certify that she had.
9.13 Decided Opinion, No. 22-2057 (2d Cir.), 91 F.4th 610 U.S. Court of Appeals for the Second Circuit (the court's own copy, as the Internet Archive captured it on January 30, 2024) · January 30, 2024 · as of October 16, 2026
Rule 11: a federal court rule. By filing a paper, a lawyer “certifies” that, after “an inquiry reasonable under the circumstances,” its legal arguments “are warranted by existing law or by a nonfrivolous argument” for changing it, the Second Circuit explained in 2024.
9.14 Decided Opinion, No. 22-2057 (2d Cir.), 91 F.4th 610 U.S. Court of Appeals for the Second Circuit (the court's own copy, as the Internet Archive captured it on January 30, 2024) · January 30, 2024 · as of October 16, 2026
Sanction: what a court may impose for a violation of its rules. Under Rule 11, the New York judge noted, it may range from “an admonition, reprimand, or censure” to “ordering a fine payable to the court” or “referring the matter to disciplinary authorities.”
9.15 Decided Opinion and Order on Sanctions, No. 22-cv-1461 (PKC) (S.D.N.Y.) (Dkt. 54), the case against Avianca, Inc. U.S. District Court, Southern District of New York (copy in CourtListener's RECAP archive) · June 22, 2023 · as of October 16, 2026
Told?
Decided the chatbot said it had supplied “real” authorities
Reported OpenAI’s 2023 notice: “may produce inaccurate information”
Reported Pew: 76% say it matters to be able to tell
The fine print
In a 2023 New York federal case, the court’s opinion records, the lawyer who did the research questioned the chatbot about its work, and it “responded that it had supplied ‘real’ authorities.” The court found his later explanations “shifting and contradictory,” supporting “a finding of subjective bad faith.”
9.16 Decided Opinion and Order on Sanctions, No. 22-cv-1461 (PKC) (S.D.N.Y.) (Dkt. 54), the case against Avianca, Inc. U.S. District Court, Southern District of New York (copy in CourtListener's RECAP archive) · June 22, 2023 · as of October 16, 2026
In a defamation suit against OpenAI, a Georgia court’s order records the terms of use a journalist accepted in 2023: using the chatbot “may in some situations result in incorrect Output that does not accurately reflect real people, places, or facts.” An on-screen notice said it “may produce inaccurate information.”
9.17 Reported Order Granting Summary Judgment in Favor of Defendant OpenAI, L.L.C., Civil Action No. 23-A-04860-2 (Superior Court of Gwinnett County, Georgia) Superior Court of Gwinnett County, Georgia (copy published by Reason) · May 19, 2025 · as of October 17, 2026
Asked in June 2025 how important it is “to be able to tell if pictures, videos and text were made by AI or people,” 76% of U.S. adults said extremely or very important, Pew reported. But 53% were “not too or not at all confident” they could tell.
9.18 Reported How Americans View AI and Its Impact on People and Society Pew Research Center · September 17, 2025 · as of October 16, 2026
In January 2024 a federal appeals court, the Second Circuit, said a lawyer had cited a state court decision that does not exist. In the court’s words, she reported that she “relied on a generative artificial intelligence tool, ChatGPT, to identify precedent,” and “did not read or otherwise confirm the validity” of it.
9.19 Reported Opinion, No. 22-2057 (2d Cir.), 91 F.4th 610 U.S. Court of Appeals for the Second Circuit (the court's own copy, as the Internet Archive captured it on January 30, 2024) · January 30, 2024 · as of October 16, 2026
Futurism reported on November 27, 2023, that Sports Illustrated published content “from nonexistent writers with AI-generated headshots,” whose profiles vanished after it asked the publisher, The Arena Group. Citing its “initial investigation,” the publisher called the claim of AI-generated articles “not accurate.”
9.21 Reported Sports Illustrated Published Articles by Fake, AI-Generated Writers Futurism · November 27, 2023 · as of October 16, 2026
Reuters reported on February 8, 2023, that it had found an error in Google’s own advertisement for its new chatbot, Bard: an inaccurate answer “about which satellite first took pictures of a planet outside the Earth’s solar system.”
9.22 Reported Alphabet shares dive after Google AI chatbot Bard flubs answer in ad Reuters · February 8, 2023 · as of October 16, 2026
In an August 2024 Pew survey of U.S. adults, 66% were “highly concerned about people getting inaccurate information from AI.” In a separate survey of 1,013 AI experts, whose answers Pew says represent only those who responded, 70% said the same.
9.23 Reported How the U.S. Public and AI Experts View Artificial Intelligence Pew Research Center · April 3, 2025 · as of October 17, 2026
Shared?
Decided a judge: “a gatekeeping role on attorneys”
Decided Georgia trial court: summary judgment for OpenAI, “in full”
Decided the court: users were “repeatedly warned”
The fine print
The New York judge wrote in June 2023 that “there is nothing inherently improper about using a reliable artificial intelligence tool for assistance. But existing rules impose a gatekeeping role on attorneys to ensure the accuracy of their filings.”
9.24 Decided Opinion and Order on Sanctions, No. 22-cv-1461 (PKC) (S.D.N.Y.) (Dkt. 54), the case against Avianca, Inc. U.S. District Court, Southern District of New York (copy in CourtListener's RECAP archive) · June 22, 2023 · as of October 16, 2026
On May 19, 2025, a Georgia trial court granted OpenAI summary judgment “in full” in a radio host’s defamation suit over a false answer about him, for “three independent reasons”: in context, no reasonable reader could take the output as “actual facts,” there was no evidence of fault, and no damages could be recovered.
9.25 Decided Order Granting Summary Judgment in Favor of Defendant OpenAI, L.L.C., Civil Action No. 23-A-04860-2 (Superior Court of Gwinnett County, Georgia) Superior Court of Gwinnett County, Georgia (copy published by Reason) · May 19, 2025 · as of October 16, 2026
The Georgia court reasoned that users, the journalist who saw the answer among them, were “repeatedly warned,” including in the terms of use, that the chatbot “can and does sometimes provide factually inaccurate information,” and that no reasonable reader in his position would take its output as “actual facts.”
9.26 Decided Order Granting Summary Judgment in Favor of Defendant OpenAI, L.L.C., Civil Action No. 23-A-04860-2 (Superior Court of Gwinnett County, Georgia) Superior Court of Gwinnett County, Georgia (copy published by Reason) · May 19, 2025 · as of October 16, 2026
The radio host’s lawyer argued that “a prudent man would take care not to unleash a system on the public that makes up random false statements about others.” The court was “not persuaded”: such a rule “would impose a standard of strict liability, not negligence,” liability “without fault.”
9.27 Decided Order Granting Summary Judgment in Favor of Defendant OpenAI, L.L.C., Civil Action No. 23-A-04860-2 (Superior Court of Gwinnett County, Georgia) Superior Court of Gwinnett County, Georgia (copy published by Reason) · May 19, 2025 · as of October 16, 2026
Several courts had proposed or enacted rules on the use of AI before them, the Second Circuit noted, but “such a rule is not necessary” to tell a licensed attorney “that she must ensure that her submissions to the Court are accurate.”
9.28 Decided Opinion, No. 22-2057 (2d Cir.), 91 F.4th 610 U.S. Court of Appeals for the Second Circuit (the court's own copy, as the Internet Archive captured it on January 30, 2024) · January 30, 2024 · as of October 16, 2026
The Missouri court held people without lawyers to the same duty. It urged “all parties practicing before this Court, barred and self-represented alike,” to know that “we are aware of the issue and will not permit fraud on this Court in violation of our rules.”
9.29 Decided Opinion, No. ED111172 (Missouri Court of Appeals, Eastern District) Missouri Court of Appeals, Eastern District · February 13, 2024 · as of October 16, 2026
In a July 2024 ethics opinion on its Model Rules, the American Bar Association’s ethics committee wrote that for lawyers using generative AI, duties to the court “require lawyers, before submitting materials to a court, to review these outputs, including analysis and citations to authority, and to correct errors.”
9.30 Reported Formal Opinion 512, Generative Artificial Intelligence Tools American Bar Association, Standing Committee on Ethics and Professional Responsibility · July 29, 2024 · as of October 16, 2026
Summary judgment: in the Georgia court’s words, it “is appropriate if the moving party demonstrates ‘that there is no genuine issue as to any material fact and that the moving party is entitled to a judgment as a matter of law.’” The other side must point to evidence “giving rise to a triable issue.”
9.31 Decided Order Granting Summary Judgment in Favor of Defendant OpenAI, L.L.C., Civil Action No. 23-A-04860-2 (Superior Court of Gwinnett County, Georgia) Superior Court of Gwinnett County, Georgia (copy published by Reason) · May 19, 2025 · as of October 16, 2026
Defamation: a person suing for it in Georgia must show, the court wrote, that the statements could be “reasonably understood as describing actual facts about the plaintiff,” and that they were published with “at least ordinary negligence.”
9.32 Decided Order Granting Summary Judgment in Favor of Defendant OpenAI, L.L.C., Civil Action No. 23-A-04860-2 (Superior Court of Gwinnett County, Georgia) Superior Court of Gwinnett County, Georgia (copy published by Reason) · May 19, 2025 · as of October 16, 2026
3 of 7 The number
In February 2023, researchers asked a chatbot 20 medical questions and checked up to three of the references it gave for each answer, 59 in all. How many did they find were fabricated? Guess first.
You guessed
Reported a 2023 study: 41 of 59 references fabricated
Reported a 2023 study: 55% of GPT-3.5’s citations, 18% of GPT-4’s
The fine print
Researchers asked ChatGPT (version 3.5) 20 medical questions in February 2023 and checked up to three references per answer, 59 in all. They reported that 41, 69%, “were fabricated, although they appeared real.” The other 18 were real: 11 articles (3 with minor, 5 with major citation errors), 5 websites and 2 books.
9.33 Reported Learning to Fake It: Limited Responses and Fabricated References Provided by ChatGPT for Medical Questions Mayo Clinic Proceedings: Digital Health (Elsevier); copy in PubMed Central · June 12, 2023 · as of October 16, 2026
In a 2023 Scientific Reports study, ChatGPT wrote 84 short literature reviews in April 2023, citing 636 works: “55% of the GPT-3.5 citations but just 18% of the GPT-4 citations are fabricated” (of 222 and 414 citations). The authors’ verdict: “Although GPT-4 is a major improvement over GPT-3.5, problems remain.”
9.34 Reported Fabrication and errors in the bibliographic citations generated by ChatGPT Scientific Reports (Springer Nature); copy in PubMed Central · September 7, 2023 · as of October 16, 2026
In a 2023 JAMA Oncology study, oncologists scored 104 ChatGPT answers against 2021 cancer guidelines. All 102 answers that recommended something included at least one the guidelines list; 35 also gave one they do not. In 13 of 104 (12.5%), responses were “hallucinated,” “not part of any recommended treatment.”
9.35 Reported Use of Artificial Intelligence Chatbots for Cancer Treatment Information JAMA Oncology (American Medical Association); copy in PubMed Central · August 24, 2023 · as of October 16, 2026
4 of 7 Their side
The companies’ own words, in their posts, reports and statements and as reporters printed them, and one company’s case as a court recounted it.
Reported Google: AI Overviews “generally don’t ‘hallucinate’”
Reported Google: odd overviews “certainly did show up”
Reported Google: accuracy “on par with” featured snippets
The fine print
Google’s account: AI Overviews “generally don’t ‘hallucinate’ or make things up in the ways that other LLM products might.” When they are wrong, it said, it is “usually for other reasons: misinterpreting queries, misinterpreting a nuance of language on the web, or not having a lot of great information available.”
9.36 Reported AI Overviews: About last week Google · May 30, 2024 · as of October 16, 2026
Google wrote on May 30, 2024, that people had shared “some odd and erroneous overviews” from its AI answers in Search, “along with a very large number of faked screenshots.” But “some odd, inaccurate or unhelpful AI Overviews certainly did show up,” it said, “generally for queries that people don’t commonly do.”
9.37 Reported AI Overviews: About last week Google · May 30, 2024 · as of October 16, 2026
Google also said its tests showed AI Overviews’ accuracy rate to be “on par with” another Search feature, featured snippets, and that it found a content policy violation “on less than one in every 7 million unique queries on which AI Overviews appeared.”
9.38 Reported AI Overviews: About last week Google · May 30, 2024 · as of October 16, 2026
LLM: short for “large language models (LLMs),” as the federal standards agency NIST writes it in its profile of generative AI.
9.57 Reported Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile (NIST AI 600-1) National Institute of Standards and Technology (NIST), U.S. Department of Commerce · July 26, 2024 · as of October 16, 2026
Google’s reply, as Reuters quoted a spokesperson: “This highlights the importance of a rigorous testing process, something that we’re kicking off this week with our Trusted Tester program.”
9.39 Reported Alphabet shares dive after Google AI chatbot Bard flubs answer in ad Reuters · February 8, 2023 · as of October 16, 2026
Reported OpenAI, as the court recounts: “reasonable care”
The fine print
OpenAI’s side, as the Georgia court recounts it: its expert testified, unrebutted, that it “exercised reasonable care” through efforts to “reduce the likelihood of hallucination” and by “providing robust and recurrent warnings to users about the possibility of hallucinations.”
9.40 Reported Order Granting Summary Judgment in Favor of Defendant OpenAI, L.L.C., Civil Action No. 23-A-04860-2 (Superior Court of Gwinnett County, Georgia) Superior Court of Gwinnett County, Georgia (copy published by Reason) · May 19, 2025 · as of October 16, 2026
Reported OpenAI: five covert influence operations disrupted
Reported OpenAI’s own limit: “This is not to say”
The fine print
OpenAI reported on May 30, 2024, that in three months it disrupted five covert influence operations, secret attempts to “manipulate public opinion” that “sought to use our models.” As of May 2024, it said, they did not seem to have “meaningfully increased their audience engagement or reach as a result of our services.”
9.41 Reported Disrupting deceptive uses of AI by covert influence operations OpenAI · May 30, 2024 · as of October 16, 2026
OpenAI limited its own claim: “This is not to say that these operations could not reach the higher categories in the future, especially if they are unwittingly amplified by media, influencers or politicians.”
9.42 Reported AI and Covert Influence Operations: Latest Trends OpenAI · May 30, 2024 · as of October 16, 2026
OpenAI rated the five operations on the Breakout Scale, which runs from 1 (lowest) to 6 (highest): “none of the five operations included in our case studies scored higher than a 2,” meaning activity on several platforms “but no breakout to authentic audiences.”
9.43 Reported AI and Covert Influence Operations: Latest Trends OpenAI · May 30, 2024 · as of October 16, 2026
Reported the publisher: “this is not accurate”
The fine print
The Arena Group’s statement, as Futurism printed it: “According to our initial investigation, this is not accurate. The articles in question were product reviews and were licensed content from an external, third-party company, AdVon Commerce,” which had “assured us” that humans wrote and edited them.
9.44 Reported Sports Illustrated Published Articles by Fake, AI-Generated Writers Futurism · November 27, 2023 · as of October 16, 2026
5 of 7 What fair looks like
What has been ruled, ordered, agreed and changed so far. Each line carries the exact word for where it stands.
Decided FTC rule, 2024: “such as AI-generated fake reviews”
Decided in force since October 21, 2024
The fine print
On August 14, 2024, the Federal Trade Commission announced a final rule against fake reviews, approved 5-0. It covers reviews and testimonials that “misrepresent that they are by someone who does not exist, such as AI-generated fake reviews,” and lets the agency “seek civil penalties against knowing violators.”
9.45 Decided Federal Trade Commission Announces Final Rule Banning Fake Reviews and Testimonials Federal Trade Commission (read in the Internet Archive’s copy) · August 14, 2024 · as of October 16, 2026
The rule, 16 C.F.R. Part 465, took effect on October 21, 2024. Under it, a business commits “an unfair or deceptive act or practice” if it writes, creates or sells a review or testimonial that materially misrepresents “that the reviewer or testimonialist exists.”
9.46 Decided Trade Regulation Rule on the Use of Consumer Reviews and Testimonials, 16 CFR Part 465, final rule (89 FR 68034) Federal Trade Commission, in the Federal Register (copy on govinfo.gov) · August 22, 2024 · as of October 16, 2026
On January 17, 2025, the FTC raised its maximum civil penalty for a “knowing violation of rule respecting unfair or deceptive acts or practices” to $53,088, from $51,744, adjusting for inflation as federal law directs agencies to do “every January.”
9.47 Decided Adjustments to Civil Penalty Amounts, 16 CFR Part 1 (FR Doc. 2025-01361) Federal Trade Commission, in the Federal Register (copy on govinfo.gov) · January 17, 2025 · as of October 16, 2026
Settled Lingo Telecom: $1,000,000 over its caller ID attestations
Settled FCC: 9,581 calls with an AI “deepfake message”
Settled Lingo: “A-Level attestations” on 3,978 calls
Settled Lingo must get “independent corroborating records”
The fine print
In May 2024 the FCC proposed a $2,000,000 penalty against Lingo Telecom over its caller ID attestations. On August 21, 2024, it settled: the carrier agreed to “pay a $1,000,000 civil penalty” to the U.S. Treasury and to “implement a robust compliance plan,” whose requirements run for 36 months unless extended.
9.48 Settled Order and Consent Decree, In the Matter of Lingo Telecom, LLC, File No. EB-TCD-24-00036425 (DA 24-790) Federal Communications Commission, Enforcement Bureau · August 21, 2024 · as of October 17, 2026
On January 21, 2024, two days before New Hampshire’s presidential primary, potential voters received 9,581 calls “carrying a prerecorded and generative AI deepfake message that encouraged potential voters to abstain from voting in the primary election,” the FCC’s consent decree with a carrier says.
9.04 Settled Order and Consent Decree, In the Matter of Lingo Telecom, LLC, File No. EB-TCD-24-00036425 (DA 24-790) Federal Communications Commission, Enforcement Bureau · August 21, 2024 · as of October 16, 2026
Lingo Telecom completed 3,978 of the calls with “A-Level attestations,” a caller ID trust label. It “admits,” for the decree, that beyond its own past checks and its customer’s certification and record it “took no additional steps” to check if that customer’s clients “could legitimately use the telephone number.”
9.49 Settled Order and Consent Decree, In the Matter of Lingo Telecom, LLC, File No. EB-TCD-24-00036425 (DA 24-790) Federal Communications Commission, Enforcement Bureau · August 21, 2024 · as of October 17, 2026
Among the compliance terms, the FCC said, Lingo Telecom must give an A-level attestation, “the highest level of trust attributed to a phone number,” only to calls where it supplied the caller ID number, and must verify customers and upstream providers “by obtaining independent corroborating records.”
9.50 Settled FCC Settles Case Against Provider That Transmitted Spoofed AI-Generated Robocalls for Election Interference in New Hampshire Federal Communications Commission · August 21, 2024 · as of October 16, 2026
Consent decree: a settlement that works as an order. The one between the FCC and Lingo Telecom was entered “for the purpose of terminating” the FCC’s investigation, and from its effective date it has “the same force and effect as any other order of the Commission.”
9.05 Settled Order and Consent Decree, In the Matter of Lingo Telecom, LLC, File No. EB-TCD-24-00036425 (DA 24-790) Federal Communications Commission, Enforcement Bureau · August 21, 2024 · as of October 16, 2026
Spoofed: sent with a false caller ID. In the FCC’s words about the New Hampshire calls: “The calls were placed using inaccurate, or ‘spoofed,’ caller ID information.”
9.51 Settled Order and Consent Decree, In the Matter of Lingo Telecom, LLC, File No. EB-TCD-24-00036425 (DA 24-790) Federal Communications Commission, Enforcement Bureau · August 21, 2024 · as of October 16, 2026
Settled two calling companies: “shall not intimidate” voters
The fine print
On May 23, 2025, a federal judge entered a consent judgment in the League of Women Voters’ robocall suit. Two calling companies acknowledged a former client’s robocall “had the potential to intimidate voters in violation of” the Voting Rights Act. They “shall not intimidate” voters and agreed to a compliance plan.
9.52 Settled Consent Judgment between the plaintiffs and Life Corporation and Voice Broadcasting Corporation, Civil Action No. 1:24-cv-00073-SM-TSM (D.N.H.) (Dkt. 121), the League of Women Voters of New Hampshire’s suit U.S. District Court, District of New Hampshire (copy on the League of Women Voters' case page) · May 23, 2025 · as of October 17, 2026
Decided appeals court: lawyers must “read, and thereby confirm”
Decided letters to “each judge falsely identified”
The fine print
The Second Circuit set out the minimum: “At the very least, the duties imposed by Rule 11 require that attorneys read, and thereby confirm the existence and validity of, the legal authorities on which they rely.”
9.53 Decided Opinion, No. 22-2057 (2d Cir.), 91 F.4th 610 U.S. Court of Appeals for the Second Circuit (the court's own copy, as the Internet Archive captured it on January 30, 2024) · January 30, 2024 · as of October 16, 2026
The New York court ordered the lawyers to tell the people most affected. Within 14 days they had to mail letters to their client and to “each judge falsely identified as the author” of a fake opinion, attaching the ruling, the hearing transcript and the fake “opinion” in that judge’s name.
9.54 Decided Opinion and Order on Sanctions, No. 22-cv-1461 (PKC) (S.D.N.Y.) (Dkt. 54), the case against Avianca, Inc. U.S. District Court, Southern District of New York (copy in CourtListener's RECAP archive) · June 22, 2023 · as of October 16, 2026
Reported Google: “more than a dozen technical improvements”
The fine print
Google said it made “more than a dozen technical improvements,” among them better detection of “nonsensical queries that shouldn’t show an AI Overview,” limits on “satire and humor content,” and limits on “user-generated content in responses that could offer misleading advice.”
9.55 Reported AI Overviews: About last week Google · May 30, 2024 · as of October 16, 2026
Reported the publisher: “removing the content”
The fine print
In the same statement the publisher said it had learned that the contractor “had writers use a pen or pseudo name in certain articles to protect author privacy — actions we don’t condone,” and that it was “removing the content” and had “ended the partnership.”
9.56 Reported Sports Illustrated Published Articles by Fake, AI-Generated Writers Futurism · November 27, 2023 · as of October 16, 2026
6 of 7 The stamp
Yours to give. There is no right answer, and the museum keeps no count.
In the room, this is where you stamp your ticket with one of five stamps. The plain version asks nothing of you.
Kept on this device only. Nothing is sent.
7 of 7 Three doors
Three ways on. All three lead to the series’ own site.
From the book · The HallucinationA bibliography from the imagination department.