Published On: July 10, 2026
Kategorie: Auswahl eines Anwalts
Artificial intelligence has made legal information more accessible than ever. With a few prompts, someone can ask an AI tool whether they have viable lawsuit, have a contract drafted, summarize a court decision, or research a potential legal claim. But accessible information is not the same as reliable legal advice.
Increasingly, prospective clients arrive at our office having done their own research on the merits of their potential litigation. Their research has been created and organized by AI. Existing clients now attempt to advise us on how to manage our cases or draft documents, claiming that “AI has all the answers.” The unfortunate reality is that AI is causing nightmares for our prospective and current clients because it is often filled with consequential flaws, incorrect advice, and is based on biased or inadequate prompts, incomplete context, or information that may not reflect the current law in the relevant jurisdiction.
In this article, we will discuss the current state of AI-generated legal conclusions and the concerns that exist for both non-attorneys and attorneys. AI is constantly evolving, and attorneys will need to adapt their practices as these tools become more capable. But for people attempting to handle legal matters themselves, the fundamental problem remains: AI can provide information about the law without offering the judgment, verification, and strategic analysis required to apply that law to a specific case.
Sorry to say to most of the DIY folks out there, AI is not going to replace attorneys anytime soon.[1]
Can AI Provide Reliable Legal Advice?
AI can be useful for learning about legal concepts, generating questions for an attorney, organizing information, or helping someone understand an unfamiliar document. The problem begins when a person treats an AI-generated response as a legal conclusion rather than as a starting point for further research.
A legal answer depends on more than the question presented. It can depend on the jurisdiction, the specific facts, the procedural posture of the case, the applicable statute, controlling precedent, contractual language, and facts the person asking the question may not realize are legally significant.
Two people can ask an AI tool essentially the same question and receive different answers because small differences in the facts or the way the prompt is framed can change the analysis. That is one reason AI-generated legal research should not be treated as a substitute for advice from a qualified attorney
AI ist wie Magie
Arthur C. Clarke sagte: "Jede hinreichend fortgeschrittene Technologie ist von Magie nicht zu unterscheiden."[2] Als ich vor ein paar Jahren anfing, mit künstlicher Intelligenz herumzuspielen, fühlte es sich an, als hätte mich ein Hellseher aus dem Publikum ausgewählt, um meine Gedanken zu lesen. Ich lernte jedoch schnell, dass KI ihre Grenzen hat, vor allem, wenn es um juristische Fragen geht, und die Nachrichten über Anwälte, die im ganzen Land in Schwierigkeiten gerieten, weil sie die Technologie einsetzten, machten mir Angst.[3]
KI-Tools erscheinen zwar ausgeklügelt, verfügen aber nicht über die Überprüfungsmöglichkeiten, das strategische Urteilsvermögen und das aktuelle Rechtswissen, das eine professionelle Rechtsvertretung bietet. KI kann ein guter Ausgangspunkt für die Betrachtung und das Verständnis von Rechtskonzepten sein, aber sie wird ein professionelles juristisches Urteil, eine Überprüfung und eine Strategie nicht ersetzen. Das sage ich nicht nur, weil ich Anwalt bin. Lassen Sie uns also sehen, warum das so ist.
What Legal Information Does AI Actually Have Access To?
The important question for someone seeking legal advice is not simply whether an AI system has been trained on a large amount of information. It is whether the system has access to the specific, current, and authoritative legal sources needed to answer the question being asked.
AI is dependent primarily on (1) public internet content, rather than private or premium legal databases (e.g., Westlaw, Lexis) and (2) free legal resources, including government websites and published case law. While this is a massive amount of data, there are limitations due to time constraints (meaning the case law may not be up to date), gaps in quality control, and potential bias.[4]
Hallucinations and Fabricated Legal Authorities
Consequently, the use of AI can result in “hallucination or fabrication.”[5] Additionally, AI systems can sometimes produce citations to cases, statutes, or other legal authorities that do not support the proposition for which they are cited, or in some circumstances, authorities that do not exist at all.
That risk is particularly serious in legal research. A fabricated citation may look convincing to someone without legal training, especially when it is presented alongside a confident explanation of the law.
AI is also not inherently aware of jurisdictional boundaries. A response may discuss California law when the relevant case is governed by Colorado law, cite federal law when state law controls, or rely on an outdated rule without adequately identifying the problem.
Das Problem mit Eingabeaufforderungen
Die Ursprünge der künstlichen Intelligenz reichen bis in die 1950er Jahre zurück und waren bis vor kurzem eher ein Produkt der Science-Fiction als der Realität. Eine große Herausforderung besteht darin, die künstliche Intelligenz mit den Grenzen der Rechenleistung und der natürlichen Sprache zu verbinden[6].
Müll rein, Müll raus
Prompts, or user inputs, are crucial to the accuracy of AI. The prompt is critical because AI systems generate responses based on the information and instructions provided to them; they do not independently know which facts have been omitted or which assumptions are incorrect.
What does that mean? I have many clients come in with their AI analysis of their case. Many prompts look like this: “Can I sue for this?” This prompt, standing alone, will always come with the answer: “yes.” An AI-generated answer can only be as useful as the information and assumptions underlying the question. If legally significant facts are missing, the response may be confidently wrong without the user realizing why
Another example is when someone asks an AI to draft a “simple” contract between parties in two different countries using a translation program. These contracts are especially problematic due to nuanced language differences, jurisdictional differences, and other elements that are not addressed in the contract. The result is a defective or inaccurate contract that appears adequate to an inexperienced client but is flawed by the limitations of AI.
Es fehlt der Kontext
DIY legal research cannot interact with an attorney who can assess credibility, clarify information, and dig deeper for intentional or unintentional disregard of damaging information or exaggeration about the situation. As most attorneys know, the case your client presents to you at the beginning is their audition to “win” you as their attorney.
AI is unlikely to discern the missing pieces, and in situations involving trials before a jury or a court, it will not be able to assess how your client will fare in their case on a personal level.
It also cannot independently interview witnesses, evaluate their credibility, inspect physical evidence, observe a client’s demeanor, negotiate with opposing counsel, or make the hundreds of judgment calls that arise during litigation.
Einrahmung
How the prompt is framed will change the output (and often raises ethical issues at the same time). For instance, “How can I get out of this contract?” is a very different question from “What provisions in the contract address termination or enforceability of the contract?”
The second question is more neutral and focused, but even a well-framed prompt does not eliminate the need to verify the resulting legal analysis. The quality of the question matters, but so does the quality and completeness of the information used to answer it.
Careful framing of the question, with tools that are law-specific (Westlaw, Nexis), and the assistance of an experienced attorney will enable you to more adequately (and ethically) guide you in a contract dispute.
Der Anwaltsunterschied
A licensed attorney provides value that AI cannot replicate, including, but not limited to: applying legal rules to specific facts, identifying risks and weaknesses, and developing a strategy tailored to the client’s goals.
An attorney also takes professional responsibility for the advice provided and can adjust that advice as the facts, law, and circumstances change
Verwendung von Eingabeaufforderungen
Die juristische Ausbildung, der Hintergrund und die Erfahrung eines Anwalts ermöglichen genauere und vollständigere Eingabeaufforderungen durch das Verständnis für:
- Die Informationen, die rechtlich und faktisch für den Fall des Mandanten relevant sind.
- Wie man Fragen formuliert, um hilfreiche strategische Hinweise zu erhalten.
- Wann man Folgefragen stellen sollte, um die Antworten der KI zu klären.
- Wie man KI-generierte Informationen verifiziert und abgleicht.
- Which facts may undermine the client’s preferred interpretation of the case.
Überprüfung
Anwälte wissen, wie sie rechtliche Informationen in zuverlässigen Datenbanken (einschließlich Westlaw und/oder Lexis, staatlichen Datenbanken, proprietären Websites und anderen Ressourcen) überprüfen können, verstehen die Hierarchie der rechtlichen Autorität und können zwischen verbindlichen und überzeugenden Präzedenzfällen unterscheiden.
An attorney does not simply ask whether an AI-generated answer sounds reasonable. The attorney must determine whether the legal authority actually supports the conclusion and whether that authority applies to the client’s jurisdiction and circumstances.
Strategisches Denken - "Wissen, wann man sie hält, wissen, wann man sie aufgibt".
Die juristische Ausbildung schult das Urteilsvermögen darüber, wann man kämpfen, wann man verhandeln und wann man sich zurückziehen sollte - Entscheidungen, die ein Verständis für Konsequensen erfordern, die KI nicht vorhersehen kann.
A legally viable claim is not necessarily a claim worth pursuing. Litigation involves costs, risks, timing, business relationships, evidence, potential recovery, and the likelihood of success. An attorney can evaluate those competing considerations when determining how to proceed.
Ethische Verpflichtungen
Anwälte sind an die Regeln der beruflichen Verantwortung und an eine Versicherung gegen Kunstfehler gebunden, was eine Rechenschaftspflicht und Regressmöglichkeiten bietet, die KI-Plattformen nicht bieten können.
Attorneys also have professional obligations regarding competence, confidentiality, conflicts of interest, and the handling of client information. Those obligations are fundamentally different from relying on a consumer-facing AI platform for legal guidance.
Beziehungsmanagement
Bei Rechtsangelegenheiten geht es oft um laufende Beziehungen, institutionelles Wissen und Verhandlungsdynamik, die menschliche Einsicht und Erfahrung erfordern.
An attorney can communicate directly with the client, opposing counsel, witnesses, experts, courts, and other participants in a legal matter. That human interaction can reveal information and strategic considerations that no standalone AI response can fully capture.
Can AI Replace a Lawyer?
AI is likely to change how attorneys research, draft, organize, and manage legal work. It may make certain tasks faster and allow attorneys to devote more time to higher-level analysis and client service. But using AI to perform a task is different from transferring responsibility for a legal matter to an AI system.
The distinction matters because clients do not simply need information about the law. They need someone who can determine which law applies, evaluate the facts, identify risks, develop a strategy, and take responsibility for executing that strategy.
Die Risiken, sich bei Rechtsberatung auf KI zu verlassen
KI kann ein Ausgangspunkt für das Verständnis rechtlicher Konzepte sein, aber sie kann und sollte professionelles rechtliches Urteilsvermögen, Überprüfung und Strategie nicht ersetzen. Das Recht ist komplex, nuanciert und entwickelt sich ständig in einer Weise weiter, die menschliches Fachwissen erfordert, um sicher zu navigieren.
Use AI Legal Information as a Starting Point, Not a Final Answer
AI can be useful for preparing questions, learning basic terminology, or organizing information before speaking with an attorney. It becomes much more dangerous when a person relies on an AI-generated answer to make a decision affecting a lawsuit, contract, business, financial interest, or other significant legal right.
Bevor Sie Entscheidungen treffen, die sich auf Ihre Rechte, Ihre finanziellen Interessen oder Ihre persönliche Freiheit auswirken könnten, sollten Sie einen qualifizierten Rechtsanwalt konsultieren, der Ihnen eine verifizierte, gerichtsspezifische und strategisch fundierte Rechtsberatung bieten kann.
The cost of proper legal consultation is usually much lower than the cost of fixing mistakes made by relying on AI for legal research. As a litigator, the author can assure you that having your contract, transaction, or other legal analysis done at the outset of your endeavors will be far less than litigation, which can run in the hundreds of thousands of dollars, years of sleepless nights, and adverse outcomes. Your legal matters deserve the protection that only qualified professional representation can provide.
Contact Lantz Law for Legal Guidance
When the stakes are high, the question is not whether AI can give you an answer. The question is whether that answer is accurate, current, applicable to your jurisdiction, and appropriate for the facts of your situation. Those are questions that require professional legal judgment.
If you are facing a legal issue or are unsure about whether AI-generated information applies to your situation, contact Lantz Law for experienced, jurisdiction-specific legal guidance before making decisions that could affect your rights or interests.
ANDERE RESSOURCEN
Zusätzlich zu den im Artikel erwähnten Materialien empfiehlt der Autor diese Bücher zur weiteren Lektüre über KI und Ethik:
Paedra Boinordiris, Beth Ruddin, Peter Scott, KI für den Rest von uns (2023).
Annette Buhler, Ethische Führung im Zeitalter der KI: Das ethische Taschenmesser (2024).
HAFTUNGSAUSSCHLUSS
Die in diesem Dokument enthaltenen und auf der Website lantzlawgroup.com verfügbaren Informationen dienen ausschließlich allgemeinen Informations- und Bildungszwecken. Die Informationen stellen keine Rechtsberatung dar. Die Lektüre der Informationen in diesem Dokument oder jeglicher Materialien auf der Website lantzlawgroup.com begründet kein Anwalts-Klienten-Verhältnis zwischen Ihnen und dem/den Autor(en) oder der Lantz Law Group. Dieses Dokument darf nicht als Ersatz für eine kompetente Rechtsberatung durch einen in Ihrem Staat oder Land zugelassenen Anwalt verwendet werden. Wenn Sie rechtliche Unterstützung in Bezug auf den Inhalt dieses Dokuments benötigen, wenden Sie sich bitte an einen kompetenten und zugelassenen Anwalt, um Hilfe zu erhalten.
[1] Ken Crutchfield, KI wird Anwälte nicht ersetzen – aber sie wird ihre Arbeitsweise verändern, Above the Law, https://abovethelaw.com/2024/03/ai-wont-replace-lawyers-but-it-will-change-how-they-work/(26. März 2025).
[2] Arthur C. Clarke war ein renommierter englischer Science-Fiction-Autor, Futurist und Erfinder, der vor allem durch seinen Roman 2001: Odyssee im Weltraum und seine visionären Ideen zur Raumfahrt bekannt wurde.
[3] Debra Cassens Weiss, Konfrontiert mit KI-Halluzinationen in Gerichtsakten zeigt ein Gericht „berechtigte Nachsicht“, während ein anderes hart durchgreift, ABA Journal https://techstory.in/lawyer-fined-5500-sent-to-ai-school-for-fabricated-chatgpt-caselaw/#google_vignette (19. August 2025).
[4] Cait Levin, KI-Eingabeaufforderung-Engineering: Wie man KI-Modelle so anleitet, dass KI-Verzerrungen vermieden werden und die gewünschten Ergebnisse erzielt werden https://techstory.in/lawyer-fined-5500-sent-to-ai-school-for-fabricated-chatgpt-caselaw/#google_vignette (September 3, 2024).
[5] Was sind KI-Halluzinationen, IBM, https://www.ibm.com/think/topics/ai-hallucinations (zuletzt besucht am 25. August 2025).
[6] Tery Winograd und Fernando Flores, Computer und Kognition verstehen – eine neue Grundlage für das Design (1. Aufl. 1986, 1987)