AI is only a tool, but responsibility is to human beings. Establish nine principles such as transparency and fairness
Mail Business has recently released “MAI” (Artificial Intelligence) news and securities agents, introducing the most artificial intelligence (AI) services among major domestic media companies, and has strengthened its guidelines for using AI based on journalism. It is determined not to lose trust and fairness, which is the intrinsic value of the media, while taking the lead in converting artificial intelligence (AX).
“AI is totally untrustworthy”…Restrict the use of sole content
Article 1 of the “AI Utilization Guidelines” enacted by the Mail Business last year stipulates that “AI is completely unreliable due to technical limitations and data bias,” and mandates the approval and review of all content by desks and reporters. In particular, through practical guidelines, it prohibits the use of AI-only content in principle and stipulates that reporters’ judgments should be involved in the entire process, from article planning to writing and editing. This means that AI will be defined as an auxiliary tool, not a substitute for reporters, to protect the value of “human-centered journalism.”
Transparency, the core of AI ethics emphasized by the World Newspaper Association (WAN-IFRA), was also reflected in detail in the guidelines for using AI. In accordance with the principle of transparency in Article 5 of the AI Utilization Guidelines, the Mail Business clearly discloses this to readers when AI is largely involved in writing articles or creating images. Beyond simply ‘using AI’, it was required to reveal what data was based on if necessary. This is a measure to ensure readers’ right to know and to remove ambiguity in AI content.
The provisions to protect its identity as an economic magazine are also noticeable. Article 2 and practical guidelines caution against relying on the interpretation of AI in areas where expertise is required, such as economic indicators or industry analysis. Even if AI summarizes the data, in-depth analysis and explanation must be left to the reporter to maintain the depth of the information.
It has also raised its defenses against personal information leakage and copyright infringement, which has emerged as a new threat in the AI era. According to Article 7, the act of entering sensitive personal information or confidential information into AI is strictly prohibited, and in accordance with Article 8 copyright protection principles, the act of unauthorized learning of other people’s content is also blocked.
In addition, in order to overcome the “bias” that AI algorithms can have, the principle of fairness (Article 6) has been prepared to allow reporters to re-verify whether the article contains discrimination or hatred against certain groups. Mail Business plans to continue to upgrade its guidelines in line with future technological changes and regularize AI ethics training for all employees.
Mail Business’s policy to strengthen AI ethics is in line with global media trends. Recently, the world’s leading media outlets, including the New York Times (NYT), the Associated Press, and Reuters, have enacted strict ethical rules along with the introduction of a Generative AI. The Associated Press stated that “AI-generated results are not exported without fact-checking.” The Guardian declared, “We will seek the use of Generative AI tools in the editing process only if it contributes to the production and distribution of original journalism.”
[Full Story] Guidelines for using AI in Maeil Business Newspaper
Preface: The Responsibilities and Utilization Principles of the Media in the Age of AI
AI technology is bringing innovation, such as increasing news production efficiency, but at the same time, it poses new challenges such as fairness, accuracy, and ethics. Maeil Business Newspaper creates and complies with the following internal guidelines under the principle that AI does not replace reporters but only uses them as an auxiliary tool for fairness and professionalism, which are the core values of the media. Guidelines are reviewed and supplemented regularly.
Article 1 Human Management and Supervision
AI cannot fully trust the results because of technical limitations and bias in the data. Therefore, Mail Business reporters must review AI-made content, and the desk must approve it and take responsibility for the overall writing of the article. If AI-generated content is reflected in the article, it must be approved by the Mail Business desk, and the final responsibility lies with the Mail Business desk and the reporter.
Article 2 Accuracy and expertise
The information provided by AI is considered to be unverified, and reporter Mail Business must cross-check whether it is true. In particular, in economic reporting, the analysis and explanation of the Mail Business reporter must be essential to maintain expertise.
Article 3 Responsibility for AI content
Even if it is an article using AI, reporter Mail Business should be responsible for the contents. All articles and content using AI must comply with the editorial code and relevant laws and regulations, and if a problem occurs due to an error in AI, it should be adequately explained and responded to the reader.
Article 4 Permissible Scope and Limitations
AI can be used for auxiliary tasks such as researching data, drafting articles, translating, and summarizing. However, it restricts its use in coverage that requires fact-checking, creative writing, and image generation that can distort reality.
Article 5 Transparency
Articles or services using AI should disclose that to readers. How AI was used and what data it was based on should also be provided if necessary.
Article 6 Fairness and Prevention of Deflection
It is necessary to review whether AI products contain prejudice or discrimination against specific groups and correct them by reflecting various perspectives.
Article 7 Protection and Security of Personal Information
Sensitive personal information or confidentiality should not be entered into AI, and readers’ data should be collected to a minimum and protected thoroughly. Respond immediately if there is a security issue.
Article 8 Copyright Protection
Be careful not to plagiarize other products by AI-created content, and do not enter other people’s content into AI without permission.
Article 9 Education and Improvement
Regular and continuous internal and external training is conducted in line with the rapidly changing AI technology and journalism environment. Members of the Mail Business should actively participate in related education.
Practice Guidelines for Writing Mail Business Articles Using AI
Article 1 Human Management and Supervision
Throughout the article writing and editing process, all steps from planning essential articles to writing, editing, and inspecting are judged and conducted by reporters in Mail Business. The sentences and ideas proposed by AI are reviewed by Mail Business reporters and desks and decided whether to adopt them or not.
Restrictions on the use of AI-only content Do not use articles or edits created by AI alone. It is also prohibited to export AI products as they are without review by a Mail Business reporter. Journalists can use a large part of AI-generated content except for simple articles that deliver data-like information such as weather, sports game results, disaster breaking news, and financial indicators.
Strengthen desk approval procedures Articles with high AI contributions, including many paragraphs or images generated by AI, must go through strict description of the Mail Business desk and consult with related departments such as the Digital News Department if necessary.
Even if AI has created some of the contents, the responsibility for the contents lies with the reporter who wrote the article and the desk who inspected it, as long as the byline of the reporter Mail Business is posted in the article. Keep in mind that AI is only a tool and cannot be a responsible entity.
Article 2 Accuracy and expertise
Fact verification required: All AI-generated sentences, data analysis, graphics, etc. must be verified by the Mail Business reporter.
Filtering Errors and Disinformation: Since AI-generated content may contain incorrect information from numbers, units, names, etc., Mail Business reporters filter it by comparing it with external data.
Maintaining expertise in the economic field: Economic indicators or industry analyses require Mail Business reporters to re-examine AI’s interpretation and add their own judgments and explanations.
Article 3 Responsibility for AI content
Compliance with laws: Strictly adhere to relevant laws such as copyright law and personal information protection law in all processes of using AI. In the event of a dispute, respond in accordance with the legal advice of an Mail Business counsel.
Compliance with the editorial code: apply the same existing editorial code and ethical standards to all forms of news, including articles, images, and videos in which AI participated in the creation.
Maintaining Quality: Even if AI intervened, the core values of Mail Business – content quality, fair perspective, and economic expertise – are not blurred.
Responsibility to the reader: Responsible for explaining and responding to errors caused by AI or raising readers’ problems.
Article 4-1 (Acceptable scope of AI utilization)
Idea planning and data research: AI can be used to set the topic of an article or to check existing data.
Drafting and translating: AI can be used as a supplement for sentence drafting, title recommendation, summary, translation, typographical correction, etc.
Personalization Services: Mail Business can use AI algorithms for content recommendations or personalization news feeds to improve the reader’s experience. AI can be used for reader convenience such as customized news, voice reading, and summary news.
Article 4-2 (Restrictions on AI Utilization)
Core coverage is the human domain: In fact, the role of Mail Business reporters is important in covering facts and writing sensitive interview questions, so the use of AI is limited. For example, only human reporters can grasp the subtle nuances or contexts obtained by face-to-face with sources, so AI does not replace these parts.
Maintain creative writing: Be careful not to lose the journalist’s personality by relying on AI.
No distortion of reality: be careful not to get confused with facts even if you use AI when editing photos and videos. Synthesis, which is likely to change or distort reality, is prohibited.
Article 5 Transparency
Revealing whether AI is used: For articles with high AI utilization or multiple AI-made products inserted, it specifies that “AI tools are partially utilized,” and the image and video announce AI generation with watermarks or captions.
Data Input Disclosure: For articles highly involved in AI, such as large-scale data analysis using AI, it reveals the source and scope of what data was provided to AI.
Allow disclosure exceptions: When using AI that is not directly revealed in the article, such as brainstorming ideas, researching data, sorting articles, and spelling checks, you don’t have to disclose that.
Article 6 Fairness and Prevention of Deflection
Check AI creation: Check carefully to make sure it doesn’t contain biased or hateful expressions.
Reflecting a balanced perspective: Review whether the AI’s proposal is biased toward one side, and supplement it by additionally covering the opposite view if it is one-sided.
Correction when finding bias: Articles written by AI are re-examined by Mail Business reporters and editors to refine them into neutral and balanced expressions. Even after the article is posted, if a related problem is found, it is immediately corrected, and if biased recommendations are made in personalized news services with AI algorithms, it is corrected. If necessary, the amendments shall be reflected after review by the Mail Business Independent Committee.
Article 7 Protection and Security of Personal Information
Do not enter sensitive information: Do not enter sensitive information such as anonymous informant information into the AI system. When it is unavoidable to use technology, identifiable information should be removed or anonymized before use.
Protecting user data: Collect the minimum amount of information needed for personalized services, and allow users to choose whether to use personalized services. It also allows you to cancel or delete it at any time.
Response to security incidents: If abnormal signs such as leakage of in-house databases, attempts to steal customer information, hacking, smishing, and DDoS attacks occur, respond immediately to reduce damage.
Article 8 Copyright Protection
Check for plagiarism: Make sure AI-generated content is similar to existing works. If similar, delete or replace.
Use third-party content: Do not enter all or photos of copyrighted content from other companies, such as other media companies, into AI, and make sure to specify the source when using it.
Article 9 Education and Improvement
Regular training and training: Operate AI utilization training programs for all employees to enhance understanding of AI tool utilization, fact-checking techniques, and ethical issues. In order to increase class participation, attendance confirmation and post-class evaluation procedures are prepared, and a guidebook is provided to familiarize yourself with related contents from time to time after class.
Periodic inspection of guidelines: In accordance with AI technology development or changes in laws and systems, the guidelines are regularly reviewed to ensure that the contents of the guidelines conform to reality, and when new types of AI services emerge or ethical issues arise, the guidelines are advanced by adding or revising relevant provisions.
Reflecting practical feedback: actively collecting difficulties or suggestions experienced by reporters and editors using AI in the field into the guidelines.
link