Siddaramaiah used both social media platforms on Tuesday to pay tribute to Sarojadevi, a revered figure in Kannada and Indian cinema, known for her roles in over 200 films. However, users flagged that the Kannada message, when auto-translated by the platform’s algorithm, was rendered inaccurately in English, failing to reflect the intended solemnity of the post. The translated text was grammatically distorted and carried a tone that some readers described as disrespectful.
Reacting sharply, Siddaramaiah criticised Meta for failing to adequately support regional languages on its platforms. He argued that the error was not just a minor technical glitch but a sign of negligence towards Kannada users and the language itself. His post, written in Kannada, called out the platforms for the error and questioned whether sufficient linguistic oversight was being exercised by the company.
The backlash prompted Meta to acknowledge the issue. A spokesperson for the company confirmed that an internal error had caused the mistranslation and that it was promptly fixed once identified. They issued an apology for the incident and reassured users that steps were being taken to avoid similar errors. The spokesperson stated that the company regretted the inconvenience caused and recognised the sensitivity surrounding posts involving tributes and obituaries.
This is not the first time Meta has faced scrutiny over its handling of Indian regional languages, particularly those with complex grammar and syntactic structures like Kannada, Tamil, and Malayalam. As the platforms expand their reach into vernacular-speaking regions, challenges in localisation have led to criticism from users and political leaders alike. Language experts have often noted that while English-Hindi translations receive considerable attention, several regional languages remain poorly served by Meta’s algorithmic tools.
Linguistic activists and digital rights groups argue that underrepresentation of regional languages in content moderation and translation frameworks not only alienates a large segment of users but also reinforces systemic bias. Kannada, which is spoken by over 50 million people, has seen limited AI support in automated translation systems. Critics say that Meta’s current AI models are inadequately trained on context-sensitive phrases and fail to grasp cultural nuances, particularly in emotional or ceremonial language.
The error involving Siddaramaiah’s condolence message has reignited this debate, with state officials urging technology firms to invest more in inclusive language training for AI systems. Several tech policy researchers point out that automatic translations are often trained using limited data sets, which tend to exclude colloquial or ceremonial usage found in regional languages. As a result, such systems tend to either omit crucial emotional context or mistranslate it entirely.
Siddaramaiah, a seasoned political figure with a strong Kannada-first approach, has frequently spoken out about the importance of preserving linguistic identity in the digital era. His confrontation with Meta adds to a growing chorus of political voices demanding greater accountability from global tech firms operating in India’s multilingual ecosystem. He urged the company to introduce manual review systems for sensitive posts written in regional languages and to involve native speakers in training data curation.
Meanwhile, users from Karnataka and other southern states have expressed solidarity with the Chief Minister, noting that errors of this kind reinforce a feeling of digital exclusion. Several social media users echoed his concerns, tagging Meta and demanding improvements in Kannada language support across both Facebook and Instagram. The issue also spurred renewed discussion on social platforms about the need for broader decentralisation of AI language development.
Following the apology, Meta reiterated its commitment to improving local language experiences and mentioned that engineers were working to refine the translation model for Kannada. Although the company did not elaborate on the exact nature of the fix, industry experts speculate that it involved tweaking the neural machine translation parameters and possibly re-weighting the context recognition layer in the algorithm.