Safeguarding voters from AI abuse

It is crucial for politicians to have a ‘produced with AI assistance’ disclaimer if they draft their speeches using AI. — Others

THE public needs to be protected from any exploitation or possible misuse of content produced by artificial intelligence (AI), especially during election seasons.

Experts and politicians say the trustworthiness of such material will have to be assured through transparency in the wake of possible automated production of deepfake speeches or imagery, among others.

Such a threat is real; several social media platforms have long put in place gatekeeping measures to mitigate negative content through the use of machine learning process.

TikTok, for example, uses automated moderation technology to identify content which violates its community guidelines.

“These systems look at a variety of signals across content, including keywords, images, titles, descriptions, and audio,” TikTok says at its website.

It also has human moderators whose responsibilities include reviewing flagged content, reports from users, and trending harmful content.

It was reported in April that TikTok has updated its guidelines by including content created or modified by AI, which it classifies as “synthetic media”. According to searchenginejournal.com, creators must label such content to curb potential risks of spreading misinformation.

“To reduce potential harm, synthetic media featuring real private individuals is prohibited. Private individuals include anyone under 18 and adults who are not public figures.

The use of public figures over 18 – government officials, politicians, business leaders, and celebrities – is permitted, but with restrictions.

“Creators must not use synthetic media to violate policies against hate speech, sexual exploitation, and severe harassment. They must also clearly disclose synthetic media and manipulated content that depict realistic scenes with fake people, places, or events.”

If not labelled, any content created with synthetic media will be removed when detected, says TikTok.

As many tech experts have said, however, AI-generated text is not easy to spot.

On Friday, it was reported United States president Joe Biden announced seven of the most influential AI tech companies’ voluntary pledges – including Google and OpenAI – to put up safeguards on the use of generative AI including watermarking AI-generated content. Biden is also working to draft legislation on regulating AI, on the heels of concerns voiced by consumers and ethicists.

Meta, which owns Facebook, Instagram and WhatsApp, among others, implements several measures to curb the spread of misinformation or false content, including labelling them as such.

“We include warnings on posts that are rated false so that you can decide what to read or share.

“When third-party fact-checkers label content false, we significantly reduce its distribution so fewer people see it,” Meta says on its site.

In the same vein, former Shah Alam Umno youth chief Mohd Dendera Billah Zamzuri says the drafting of speeches through the AI-driven ChatGPT bot, for example, will need to be labelled as “produced with AI assistance” for accountability of its owner and reader upon delivery.

He says this is particularly crucial for top political leaders, whose words will dictate party or government policy.

“The trustworthiness of information or proposals produced by generative AI remains to be questioned especially during electoral campaigns. As such, speeches created by ChatGPT should be mentioned as so. This is transparency, and particularly important for party leaders.”

A political party insider echoes Mohd Dendera’s views, saying it is one way to protect the people’s interests, including party members, during election season when promises or pledges are rife.

“At the end of the day, the speeches will be seen as their organisation’s policy.”

Concurring, Muda data and technology director Faezrah Rizalman says political parties and candidates should be obliged to disclose their AI strategies and also provide clarity on how AI algorithms are employed to target voters.

“This transparency would enable scrutiny and ensure responsible and ethical AI usage.”

She stresses specific regulations and guidelines should be established to govern AI usage in electoral campaigns.

“These rules can define permissible AI applications, emphasise the importance of fair and transparent campaigning practices, and establish boundaries to prevent the dissemination of false or misleading information.

“Independent regulatory bodies could be entrusted with overseeing compliance and enforcing these guidelines. For example, regulations could address the ethical collection, storage, and analysis of voter data, drawing lessons from international best practices like the General Data Protection Regulation in the European Union.”

With the possibility of AI being used to create fake news or target voters with personalised messages that are designed to manipulate them, it remains pertinent to be aware of such risks.

The formation of ethical guidelines for using AI in electoral campaigns should be carried out by various parties, says policy and political lobbying firm Imperium Consulting’s chief executive officer Rahman Hussin.

“These guidelines should be formed by a variety of stakeholders, including political parties, civil society groups, and technology companies.”

Ultimately, it is imperative for the people to be aware of the capabilities, uses and abuses of AI, says Bar Council’s technology, cyber and privacy law committee chairman Sarah Yong Li Hsien.

“When people know that what they see or hear may not be true or real, or could be designed to manipulate them, they would be conscious that they should not receive information at face value and should always fact-check.”

Other than awareness and AI-literacy programmes, she also proposes a greater transparency on online platforms concerning curated contents and profiling.

“We would also need to advocate for more transparency from online platforms when it comes to targeted advertising and profiling, as well as the criminalisation of profiling based on certain sensitive personal data.

“One of the initiatives of the Bar Council’s technology, cyber and privacy law committee is to conduct webinars on AI for awareness, not just within the legal fraternity, but pertinently for the public at large.”