Kannada-English translation error sparks concerns over automated multilingual tools in public communications
Meta has issued an apology after a mistranslation by its auto-translation tool on Instagram incorrectly announced the death of Siddaramaiah, the Chief Minister of Karnataka, a southern Indian state. The incident has raised serious concerns about the reliability of automated translation in high-stakes political communication, especially for under-resourced languages like Kannada.
The erroneous translation stemmed from a post Siddaramaiah made in Kannada on Instagram, paying tribute to the late Indian actress B. Saroja Devi. However, the platform’s English translation read as if Siddaramaiah himself had passed away. The distorted version was reported by local media, reading: “Chief Minister Siddaramaiah passed away yesterday… took darshan of Sarojadevi’s earthly body and paid his last respects.”
Meta responded by confirming that the issue had been fixed and offered an apology. “We fixed an issue that briefly caused this inaccurate Kannada translation. We apologize that this happened,” a Meta spokesperson told Press Trust of India.
A Call for Caution and Accountability
Siddaramaiah condemned the incident as “dangerous,” calling out Meta for negligence that “can harm public understanding and trust.” His office sent an official email to Meta, urging the company to temporarily suspend its Kannada auto-translation tool until its accuracy improves significantly. The email described Meta’s system as “frequently inaccurate and, in some cases, grossly misleading.”
The message further stressed that such errors are especially problematic when dealing with public-facing communications from government officials. “Given the sensitivity of public communication, especially from a constitutional functionary like the Chief Minister, such misrepresentations due to flawed translation mechanisms are unacceptable,” the statement read.
His office also urged Meta to collaborate with Kannada language experts to refine the tool and prevent future errors.
A Broader Issue for AI Translation
Kannada, spoken by over 60 million people, remains underrepresented in many AI training datasets. This incident highlights the challenges of automated translation in lower-resourced languages, where nuance, grammar, and cultural context are often lost in current models.
As of Friday, the auto-translation on Instagram still appeared inaccurate, suggesting persistent issues with the tool’s handling of Kannada.
This is not the first time Meta has faced backlash over technical errors involving sensitive content. In recent months, the platform was also criticized for showing graphic videos to users due to a separate malfunction.
The case underscores the pressing need for transparency, linguistic diversity, and human oversight in the development and deployment of multilingual AI systems, especially when the stakes involve public officials and national discourse.
