CIJ’s response in relation to the deployment of AIFA the chatbot

The Centre for Independent Journalism (CIJ) notes the recent launch of the fact-checking chatbot named Artificial Intelligence Fact-Check Assistant (AIFA) to verify ‘fake news and information’ on WhatsApp. Communications Minister Fahmi Fadzil stated that it is part of the government’s initiative to curb misinformation and promote digital literacy among Malaysians.

We find this a positive effort to curb the proliferation and spread of disinformation, especially on private messaging platforms. It is also commendable that the chatbot can address the diversity of various languages in Malaysia—English, Malay, Mandarin, and Tamil (missing the ethnic languages in Sabah and Sarawak).

Notwithstanding, it is critical to consider the following:

1) Government as the sole arbiter of truth - The threshold of disinformation must be clearly defined to avoid the government possibly misusing it to control the prevailing narratives. Clear safeguards will be necessary to prevent the government (especially MCMC) from being the sole arbiter of what is the truth. Pro-government bias in the development and deployment of the system would undermine public trust. We require further information on how the chatbot would be moderated and which guidelines would it adhere to, especially in guaranteeing ethical use of AI, upholding freedom of expression, and privacy and data protection.

2) Speed of amplification – The dynamic nature and speed of the dissemination of (dis)information would require the AIFA to provide real-time verification. While being proactive (as delays would mean the disinformation would have been spread and the harm amplified), the chatbot must act in adherence to fundamental principles of freedom of expression and not undermine the accuracy and ethics of its response.

3) Accuracy – Given the dynamics of our linguistic diversity it would require the chatbot to be adept at discerning nuanced language contexts as the datasets using coded language, euphemisms, local slang, and use of visuals or emojis that evolve quickly with social and cultural shifts may be misinterpreted. It is critical to ensure it is accurate and to avoid misclassification of legitimate information or news as false. It would also require advanced technology to distinguish deepfakes and malfinformation.

4) Digital literacy – It would require robust public education campaigns and digital literacy programmes to encourage the public to use and integrate habitual fact-checking in all their communications. This is especially so for those who are less tech-savvy or have accessibility challenges.

5) Data protection – Another key concern is the fact that the government is exempted from the scope of the Personal Data Protection Act 2010. As such, we would require further guarantees that the data and personal information of the users will be protected. We would also require further disclosures and transparency on the obligations and accountability measures to hold the developers, deployers, and those managing data storage to account to prevent data leaks and misuse of user data.

Thus, we recommend the following to be implemented by the government:

1) To institute a clear human oversight over the deployment of the chatbot. It should also have an independent and multistakeholder oversight body which could also function as an independent dispute resolution mechanism;

2) There should be transparency of algorithms/datasets;

3) To implement regular due diligence and independent audits to identify biases and its impact on human rights;

4) To conduct regular digital literacy education and public awareness campaigns.

28 January 2025 Wathshlah G. Naidu CIJ Executive Director

The press statement is issued by the Centre for Independent Journalism (CIJ)