Could a single copyright notice one day shut down internet access in its entirety to an entire university, or hospital, or city? This is the possible future if the Supreme Court decides in favor of the country’s biggest record labels in the battle between internet service providers and companies like Sony Music Entertainment in the case of Cox Communications. The issue revolves around the issue of internet service provider liability regarding piracy done through peer-to-peer networks such as BitTorrent because this file-sharing system aims to decentralize file transfer through the division of data or files into ‘tiny’ pieces that are distributed via many ‘peers.’ This makes tracing piracy difficult because ‘a copyright holder may use piracy detection software that traces the IP address of suspected pirates and notifies the ISP.’
The record labels claim that Cox knowingly allowed “habitual offenders” to continue violating copyrights because 619,711 subscribers for nonpayment but only 32 accounts were terminated for repeat infringement during the same time period. The labels argue this made the Digital Millennium Copyright Act’s repeat infringer provision worthless because of the lack of enforcement by the company. The defense argues that infringing notices are simply accusations and that terminating accounts based upon those notices could cause “mass evictions from the internet.” This would pertain to entire households both in military barracks and in hotels.
The implications are not limited to the issue of music piracy. Large corporations such as Google and X are among those that have signed amicus briefs in support of ISPs on the basis that the ruling in the Fourth Circuit may cause “havoc” in the industry of technological innovation, including in the provision of artificial intelligence systems. It would be possible that corporations may be compelled either to limit the power of their algorithms or use filtration in excessive measure in order not to be held liable. This assumes significance in view of the fact that AI data training involves the use of voluminous data including copyrighted material that may be qualified under fair use inasmuch as the data is reduced to mathematical patterns.
However, there are technical challenges in this rationale. Machine learning models are essentially large-scale data compression algorithms that compress rules of language and structure from the data they are trained upon. This has enabled the AI models to replicate styles or reproduce text data within margins of error, leading debate over intellectual properties in this context. Cases like those brought by the artistic community against Stability AI over image scraping or The New York Times suit against OpenAI and Microsoft over news article data intake reflect the tensions between innovation and intellectual properties.
The legal context involves precedence involving secondary liability. The Betamax decision in 1984 protected Sony Corp. from liability in contributing infringement regarding the sale of VCRs, holding that operators lacking substantial infringing use should not be barred from market access. Twenty years later in the case of MGM v. Grokster, the Court ruled in favor of those seen actively encouraging the crime. The current precedent involves refusal from the Court to consider Twitter and other platforms liable regarding support of terrorists in the absence of facility action. The current case presents challenges in classifying ISPs somewhere in between.
Justices queried both extremes during oral arguments. For instance, Justice Sonia Sotomayor queried why Cablevision could not simply prevent service to individual households practicing infringement in response to Justice Alito’s inquiry about why Cablevision would be required to shut down entire schools based upon the wrongdoing of one infringer. Justice Neil Gorsuch pointed towards a ‘narrow holding’ referring the case back with a different standard based upon the “murky” current standard of secondary liability. The Deputy SolicitorGeneral of the U.S. advocated harmonization with patent laws’ codified standard requiring purposeful facilitation.
From the engineering point of view, enforcing strict disconnection notices would entail the use of more detailed traffic analysis and identification tools that may include deep packet inspection or behavioral analysis. This approach may infringe the privacy guidelines and net neutrality rules because it would be necessary to distinguish lawful from unlawful traffic in real-time. For AI platforms, this would mean ensuring that there are detailed training datasets that are licensed in particular ways or that include watermarking of derivatives in order to track downstream works, possibly impacting developer productivity.
The spillovers from this ruling may shape the future of balancing innovation and enforcement by internet infrastructure companies and AI developers. The hegemony of strict liability may coerce ISPs into being the “internet police,” or the use of more restrictive liability may render rightsholders ineffective in combating mass infringement acts. However, this verdict would most likely become precedent in future controversies over copyrights in the context of the AI world in which lines separating human imagination, machine learning algorithms, and copyrights are no more clear-cut.
