Poll overseers, get your AI act together

Google is taking the first step on disclosure of the use of AI tools and content in political advertising on its platform. This will have a global impact and is in addition to its policy about misrepresentation. Facebook has similar policies on deepfakes - synthetic images and videos created to mislead - but has not yet come out with a requirement to badge AI content. Social media platforms, which will carry most of the content created through machine learning (ML), are timing remedial measures to coincide with a clutch of national elections next year, including those in the US and India. But this will not be sufficient. Election overseers also have to get their AI act together so that the technology does not get smeared by political shenanigans.

Lawmakers need to put up guard rails for AI catering to its use across industries. Drug regulators, for instance, can approve locked algorithms that do not change over time unless updated by the developer. Financial profiling for creditworthiness by banks using ML, likewise, would require regulatory disclosure on the robustness of data-fed models. Self-censorship may possibly be the best regulatory approach to handling AI created content in media. Industry-specific regulation will co-evolve with deployment of AI.

Beyond this, of course, there is a need for a generalised system of governance that incorporates privacy protection, ethical deployment of AI, transparency of process and accountability for errors in machine-assisted decisions. A sector-specific approach may be more expedient, but will lack holistic regulation of the technology and its successors. Lawmakers were behind the curve in privacy protection. They can't afford to repeat their mistakes with AI.