Narratives
Narratives are organized along two axes: the shifting of responsibility and the question of remuneration. The hallucination and judicial AI narratives reinforce each other, as both claim that professional responsibility cannot be outsourced to machines; however, this conflicts with the efficiency promise of the platform war and the narrative regarding the end of hourly rates.
Full summary
Copyright narratives (US litigation and European forums) proceed side by side but with different logic: in the US, market harm and the licensing market are decisive, while in Europe, output and real-time retrieval are key, and the Hungarian Like Company case reinforces the latter path. The AI Act narrative is more durable than fear of labeling fines, while the chatbot and agent liability narratives raise the same product liability question regarding different forms of harm. The main tension in the field is that accelerating AI adoption continuously shifts responsibility to the user, the lawyer, the judge, and the operator, while the legal foundations of manufacturer liability (fair use, product versus service, freedom of speech) remain undecided.
A hallucinated citation is the lawyer's responsibility
An AI-fabricated citation is not a technical error but a professional breach of duty, punishable by fines, removal from the case, and disbarment, because responsibility always lies with the signing lawyer.
↗ rising · evidence: moderateJudicial AI use also requires oversight
Hallucinations have already appeared in court rulings; therefore, courts need their own strict AI rules, as decision-making cannot be outsourced to machines.
▲ at peak · evidence: moderateTraining is not free: a wave of lawsuits and the licensing market
AI companies deliberately trained on pirated sources and must pay; courts increasingly recognize that training data has a market and a price.
↗ rising · evidence: moderateEurope as the judicial forum for rights holders
Where US fair use is uncertain, European courts may treat training, output and real-time retrieval as infringement, and classify memorization as copyright-relevant. The Court of Justice of the European Union will decide these questions.
→ lasting · evidence: moderateThe AI Act is delayed but already biting
The Digital Omnibus only extended deadlines for high-risk items; transparency and GPAI rules remain in force, and the AI Office is already requesting information from companies, so compliance cannot be postponed.
↘ fading · evidence: weakLabel everything or face fines worth billions
From August 2, all AI content and chatbots must be labeled; companies are unprepared, and incorrect labeling can result in fines in the billions.
▲ at peak · evidence: moderatePlatform war for the entry point to legal work
Legal work will take place on platforms like Harvey, Legora, Thomson Reuters, and now frontier labs, requiring firms to adopt firm-level AI strategies and partnerships.
↗ rising · evidence: weakAI erodes billable hours
Time saved by AI pressures the hourly billing model: clients expect fee reductions, and AI-native firms positioned against the hourly model are launching.
↗ rising · evidence: weakThe junior lawyer's career ladder disappears
AI is taking over routine work of junior lawyers, eliminating the learning path; legal faculties and firms must redesign training.
↗ rising · evidence: moderateWho is liable if an AI agent causes damage
Autonomous AI agents are opening a liability gap. The suitability of product liability rules is debated, with little appellate case law, leading to US federal bill proposals and new product liability rules suggested by EU legislators.
▲ at peak · evidence: moderateThe chatbot is a product; its maker is liable for the harm
According to lawsuits, chatbots contributed to deaths, self-harm, and violence; therefore, their manufacturers must be held liable on product liability and consumer protection grounds, as several US states and one Canadian province have already attempted.
↗ rising · evidence: moderateThe face and voice as new property
Against deepfakes, people must be granted a property-like right over their own faces and voices, backed by criminal prohibitions and swift platform obligations.
Who is affected by which narrative
How it was made
Narrative: a recurring interpretive frame that the press, companies and the profession repeat across several events; not a single event. Machine analysis identified them from the period's 725 stories, each backed by at least six stories. A separate verification pass checked every claim back against the supporting stories and corrected or deleted any that diverged (110 corrections). The phase shows where the narrative stands at the end of the period; the evidence level shows whether it rests on measured data and cases, or only on statements and forecasts. The curve is the weekly count of supporting stories, not of the whole news flow.