Showing the incident of the Lt. Col. Stanislav Petrov in the Cold War era as the case exemplifying the importance of the “meaningful human oversight,” the author emphasizes its importance in the recent AI usage in the areas of the “high risk” or “sensitive domain” in light of “automation bias” due to which human beings are inclined to over-trust AI’s forecasts, decisions, or recommendations in spite of adoption of the “human-in-the-loop” system. Then, he introduces examples of the countermeasures against the automation bias in AI usage; such examples include the ones suggested in the Detailed Explanation on Key Points concerning AI Utilization Principles of the MIC (Ministry of Internal Affairs and Communications) in Japan, the AI ACT of EU in Europe, and the AI Bill of Rights Blueprint of the White House in the U.S. Thereafter, the author introduces human beings’ under-trust of AI, which is called “algorithm aversion,” along with its relevant allegations against the bias preferring human decisions to AI forecasts. Then, he criticizes all of them based upon:

(i) logical reasons such as the limits of AI’s ability due to which AI should not be applicable to decisions in complicated and context-dependent tasks, compared with simplistic, repetitive, and mechanical tasks; and

(ii) ethical reasons such as human dignity, respect for human beings, and necessity of taking into consideration individual personhood.

This paper examines the issues related to copyright content moderation on content-sharing platforms. Specifically, it discusses the EU's Digital Single Market Copyright Directive (DSMCD), relevant court decisions (C-401/19), and the transposition in Germany, the Act on the Copyright Liability of Online Content Sharing Service Provider (UrhDaG), focusing on the importance of regulations that consider user interests.

Article 17(4) of the DSMCD obliges online content-sharing service providers (OSCCPs) to use their best efforts to ensure the unavailability of copyrighted works, and to fulfil this obligation, OSCCPs would inevitably introduce so-called automated upload filters. However, concerns were raised in relation to users' freedom of expression and information, as upload filters may involve over-blocking; the interpretation of Article 17 given in C-490/19, suggests that other provisions of Article 17, such as Article 17(7) and (9), may take measures to prevent blocking of lawful content and to establish effective and prompt complaint procedures to challenge unjustified blocking, and that the provisions of Article 17 as a whole should be balanced, taking into account not only the interests of rights holders and service providers, but also the interests of users.

The UrhDaG has gone one step further with a policy of protection of user interests: Article 9 of the UrhDaG introduces a new category of "uses presumably authorized by law" (uses permissible under any statutory limitation to copyright) that an OCSSP must, in principle, communicate to the public.

In the face of the inevitable use of automatic upload filters, this paper shows that proper content moderation will require (1) considerations of policies for the appropriate design of upload filters, (2) enforcement against OSCCPs, (3) transparency and data access for evaluating upload filters, and (4) the content and legal nature of ex-post remedy procedures (e.g. from the perspective of 'user rights'). It concluded that the discussions in the EU have certain implications for Japanese law and that further study is needed in the future.