Negative emotions are continuing to ferment. Many countries plan to strengthen the supervision of ChatGPT.

　　The picture shows that on November 7, 2022, in the artificial intelligence area of the 5th China International Import Expo(CIIE) Technology and Equipment Exhibition Area, the experiencer felt "high-five from a distance" in the virtual workspace with the help of VR headsets. Xinhua News Agency reporter Hou Junshe

　　□ Rule of Law Daily reporter Wang Wei

　　On April 3rd, ulrich Kelber, German Federal Commissioner for Data Protection, said that Germany might follow the Italian example and temporarily ban ChatGPT for the sake of data security protection. He revealed that the act of disabling the software will fall within the purview of the German Federal Data Protection Agency. In addition, the German Federal Data Protection Agency asked Italian regulators to provide more information about its blocking of ChatGPT.

　　Experts pointed out that ChatGPT has 100 million active users within two months of its launch, making it the fastest growing consumer application in history. Although ChatGPT is amazing in practical use, it still has disadvantages such as high cost and information leakage due to the limitations of large-scale language model itself and data reasons. How to balance scientific and technological innovation and privacy protection still needs continuous exploration by all parties.

　　Multinational follow-up restrictions

　　Recently, negative emotions such as dissatisfaction, panic and conflict with ChatGPT in many countries are continuing to ferment.

　　Reuters reported that the Italian Personal Data Protection Agency (DPA) recently announced that it would ban the use of ChatGPT from March 31st, restrict OpenAI from processing Italian users’ information and data, and at the same time, initiate an investigation into its privacy and security issues. This move has attracted the attention of multinational regulators and enterprises. Germany, France, Ireland and other countries have begun to prepare to follow the example of Italy and strengthen the supervision of ChatGPT.

　　On April 3rd, German Business Daily quoted ulrich Kelber as saying that, in principle, similar shielding actions are also possible in Germany. Germany has asked Italy for further information about its ban.

　　Meanwhile, according to Reuters, the French and Irish personal data regulators have contacted their Italian counterparts to learn more about the basis of the ban. A spokesperson for the Irish Data Protection Agency (DPC) said: "We are following up with Italian regulators to understand the basis for their actions, and we will coordinate this matter with all EU data protection agencies."

　　The Canadian Broadcasting Corporation (CBC) reported on April 4 that the Canadian federal privacy regulator announced that it had launched an investigation into OpenAI because the company was suspected of "collecting, using and disclosing personal information without consent". The report also mentioned that US President Biden and his Science and Technology Advisory Committee also discussed the "risks and opportunities" of artificial intelligence at the meeting held on the same day.

　　According to South Korea’s "National Daily" reported on April 3, a number of Korean companies, including Samsung, SK Hynix and Posco, are taking measures to restrict the use of ChatGPT to prevent employees from leaking secrets. Although LG Electronics has not issued a ban, it requires employees to be trained in "artificial intelligence and data use".

　　According to the Nikkei News Network, Japanese companies, including Softbank, Hitachi, Fujitsu, Mitsubishi UFJ Bank and sumitomo mitsui banking corporation, have begun to restrict the use of interactive artificial intelligence services such as ChatGPT in their business operations.

　　However, the privacy regulator in Sweden said that there is no plan to ban ChatGPT at present, and there is no contact with the Italian regulator. Spanish regulators said that they had not received any complaints about ChatGPT privacy leaks or other aspects, but they did not rule out the possibility of conducting an investigation in the future.

　　Suspected information leakage risk

　　This is not the first time ChatGPT has caused controversy. Since the launch of the service, some people worry that it will be used to cheat, spread false news or mislead the public.

　　ChatGPT was launched by OpenAI, an American artificial intelligence research laboratory, on November 30, 2022, and it has amazing language understanding, generation and knowledge reasoning ability. According to reports, within a few days after its launch, its registered users exceeded 1 million, and the number of active users in two months reached 100 million, which set off a hot discussion on the whole network and became the fastest growing consumer application in history, setting off a huge wave of technology in the field of artificial intelligence. However, it is undeniable that when ChatGPT is used more widely, the corresponding questions and concerns follow.

　　The industry believes that ChatGPT technology is still in the early stage, and there are many problems in itself. First of all, as a large-scale language model, ChatGPT is expensive to use. ChatGPT basic model has high training cost, difficult deployment, high cost for each call, and possible delay, which requires high engineering ability. The data shows that in order to make the language synthesis result of ChatGPT more natural and smooth, OpenAI shared 45TB of data and nearly 1 trillion words to train the model. This is roughly equivalent to 13.51 million Oxford dictionaries. The cost of training once is as high as ten million dollars.

　　Secondly, the social risks of artificial intelligence technology in information leakage will be further amplified because of the huge data scale used. As far as judicial work is concerned, during the interaction between judicial personnel and ChatGPT, ChatGPT will obtain a large number of business-related data, which can not be made public at will before the case meets the conditions of judicial disclosure. Once the data is leaked, it may lead to serious consequences. In addition, the terms of use of OpenAI do not provide any protection for the data information that users may input into ChatGPT.

　　Thirdly, the problem of low accuracy of ChatGPT cannot be ignored. Because most of ChatGPT’s answers are extracted or collected from known content, wrong or inappropriate suggestions may be given for unknown or blind areas of ChatGPT knowledge. For example, judicial staff often need to take into account "justice, law and human feelings" in handling cases, and need to consider the unity of legal effects, social effects and political effects of handling cases, which have certain humanistic thinking, and these are often lacking and insufficient in artificial intelligence. Therefore, the suggestions given by ChatGPT at this time will be biased to some extent.

　　Data security has a long way to go.

　　Regarding the risks exposed by ChatGPT, the protection of personal data in the EU is at the forefront. It is reported that the upcoming "Artificial Intelligence Act" of the European Union adopts a regulatory idea of classifying and managing artificial intelligence systems based on the degree of risk. However, generative artificial intelligence products like ChatGPT challenge the current risk classification methods.

　　In fact, as early as 2021, the European Commission proposed the draft of the Artificial Intelligence Act, but the progress of the proposal was not as smooth as expected. The core of the controversy among members of the European Parliament is whether regulation will hinder innovation, that is, how to balance data privacy and avoid stifling innovation and investment in artificial intelligence. One of the most controversial areas is deciding which artificial intelligence systems will be classified as "high risk". According to informed sources, European legislators are currently negotiating new rules that restrict "high-risk" AI tools in many aspects.

　　Not only have members of the European Parliament not yet reached an agreement on the basic principles of the proposal, but there has also been a dispute within the government about whether to supervise ChatGPT.

　　On April 2nd, matteo Salvigni, Italian Deputy Prime Minister and Minister of Infrastructure and Sustainable Transportation, publicly posted on social platforms, criticizing DPA’s decision to temporarily ban ChatGPT, saying that its blocking behavior was "excessive". He called the regulator’s behavior "too hypocritical", which seemed to have no common sense. "Almost all online services can’t avoid privacy issues". In his view, this ban may damage the country’s commercial development and technological innovation, hoping to find a solution quickly and resume the use of ChatGPT in Italy as soon as possible. A German government spokesman also believes that there is no need to ban ChatGPT.

　　Experts said that on the one hand, it is resolutely banned, and on the other hand, it is tolerant. Behind the controversy, it also shows that how to protect the privacy of citizens’ data and avoid the risk of privacy leakage will be problems that all subjects must consider and solve after artificial intelligence and large-scale coverage of social life. The "one-size-fits-all" approach that is too decisive is not worth advocating. Regulatory policies should not only have flexible space for discussion, but also allow enterprises to fully release their innovation vitality. More importantly, it is necessary to protect citizens’ personal data privacy and information security. This is not only a difficult problem facing Italy and Germany now, but all countries need to strike a good balance between protecting data security and excessive supervision and control, and use leverage to support both ends of technological innovation and privacy protection.