Reporting in Accordance with the GRI Standards遵循GRI標準寫報告書

 有次上課講師講了一件事，他説有幾家上市公司被金管會要求要重寫2024的ESG報告書，其中有一家修改好了，二個獨董卻不敢簽字，原因可能是他們也不知道資料到底對不對。

現在大家都意識到永續資訊揭露不實，可能會被罰鍰或有其他法律責任。所以ESG報告書現在也有合規的問題，除了國內法規還有報告書依循標準的問題。

一般廠商都會在報告書中說明是依照GRI標準來寫的，但是狀況不一樣，In accordance with GRI Standards是遵循GRI標準寫的，另一種是參照GRI標準Reference to GRI Standards。

依照GRI 1：Foundation 2021（2023 生效）規定，你要聲稱是否「依循」GRI標準，需滿足 幾項強制要求，包含：
1.使用聲明（in accordance statement）
2.GRI 內容索引表
3.明確使用 GRI 1、2、3
4.重大主題決定流程
5.通報 GRI（非強制但為實務常態）

多數廠商最會有問題的地方在重大主題管理，沒有按照GRI的要求做報告書：
GRI 3-3 要求你 「逐一針對每一個重大主題」說清楚三件事：
① 該重大主題「對誰造成什麼衝擊」，實際或潛在衝擊，正面或負面，衝擊對象
② 公司「如何管理這些衝擊」
政策、制度、流程、責任單位
管理方式（不是只有活動描述）
③ 公司「如何評估管理是否有效」，KPI監督機制, 查核、稽核、績效追蹤方式

多數公司在管理衝擊的方式、行動等沒有撰寫，變成你要說是In accordance with GRI Standards，又明顯有地方沒做到位。

其實現在要參照的標準挺多的，何不讓AI來幫你寫報告書？你可能會想說用AI那簡單，我也會用。

我有次讓Chatgpt 摘要一個文件，它給了一段文字後我想再去看原文，請它給我頁數，它也給了，耶 居然找不到，我再問了它幾次，它才承認錯誤，它把GRI以前在徴求外界意見，沒有定案的資料也抓進來回覆我了，所以你要知道AI給的資料可能不見得正確。

不過如果AI程式已經把GRI 1/2/3, 200/300/400條文都放裡面，把每個條文用節點隔開，避免互相干擾，條文還有對應的Excel表格，你只要把資料填好上傳雲端後，就是產出對應該條文的文字讓你使用， 這樣你比較不用擔心合規的問題。

另一方面是企業可因此讓人力資源運用更有效率，不用因為製作ESG報告書佔據ESG部門人員太多時間，更不用說企業可能精簡人力，沒有太多人力專職在做ESG相關的業務，更須要這種AI程式的協助。

During one class, the instructor mentioned something: several listed companies were required by the Financial Supervisory Commission (FSC) to rewrite their 2024 ESG reports. Among them, one company completed the revisions, but two independent directors refused to sign off—likely because even they weren’t sure whether the data was accurate or not.

Nowadays, everyone is becoming aware that inaccurate or misleading sustainability information disclosure can lead to fines or other legal liabilities. As a result, ESG reports now carry real compliance risks—not only under domestic regulations, but also in terms of the standards the report claims to follow.

Most companies state in their reports that they are prepared “in accordance with the GRI Standards.” However, there’s a key distinction:

• “In accordance with GRI Standards” means the report fully complies with the GRI requirements.
• “Reference to GRI Standards” (or “with reference to”) means the report only refers to or draws from parts of GRI without full compliance.

According to GRI 1: Foundation 2021 (effective 2023), to legitimately claim that a report is “in accordance with” the GRI Standards, an organization must meet several mandatory requirements, including:

1. An “in accordance” statement
2. A GRI content index
3. Explicit use of GRI 1, GRI 2, and GRI 3
4. A clear process for determining material topics
5. Notification to GRI (not strictly mandatory, but common practice)

The area where most companies fall short is in the management of material topics. They often fail to report in line with GRI requirements.

Specifically, GRI 3-3 requires that, for each material topic, the organization clearly explain three things: ① What impacts (actual or potential, positive or negative) the material topic causes on whom (the impacted parties) ② How the company manages those impacts—including policies, systems, processes, responsible units, and management approaches (not just descriptions of activities) ③ How the company evaluates whether its management is effective—including KPIs, monitoring mechanisms, audits, verifications, and performance tracking methods

Many companies only describe actions or events without properly addressing how they manage impacts or evaluate effectiveness. As a result, even if they claim “In accordance with GRI Standards,” there are obvious gaps where the requirements are not met.

In reality, there are now quite a few standards that companies need to reference. Why not let AI help write the report? You might think, “Using AI is simple—I can do it myself.”

I once asked ChatGPT to summarize a document. After it gave me a paragraph, I wanted to check the original text and asked for the page numbers. It provided them, but when I looked, I couldn’t find them. After several follow-up questions, it finally admitted the mistake: it had pulled in draft or consultation-version content from GRI that had not yet been finalized. This shows that the information AI provides may not always be accurate.

However, if an AI program has already incorporated the full text of GRI 1, 2, 3, and the 200/300/400 series, structured each clause as separate nodes to avoid cross-contamination, and paired them with corresponding Excel templates—then users can simply fill in their data, upload it to the cloud, and generate compliant text corresponding to each disclosure. This approach significantly reduces compliance risks.

On the other hand, companies can use this to allocate human resources more efficiently. Preparing ESG reports often consumes too much time from ESG team members. Many companies have already streamlined their workforce and lack dedicated full-time staff for ESG tasks—making tools like this kind of AI program especially valuable and necessary.