ULG's Language Services Blog

Six Risks of Free Machine Translation Tools

Free machine translation (MT) tools such as Google Translate are alluring for businesses that want to optimize their translation budget or speed up their translation processes. What could be cheaper than free? And what could be faster than copying and pasting your content and hitting the translate button?  

The hard truth is that these tools aren’t really free of cost. Whether you’re using neural machine translation (NMT) tools like DeepL, Google Translate, or Bing Translator, or an LLM like ChatGPT translation, these widely available tools often come at a price—a price your organization can't afford to pay. 

One key difference between these free tools and paid MT tools is customization. Free tools are “one size fits all,” but businesses have different translation needs than the public translating things like product reviews and other information on the fly.  

Paid machine translation software is customized to provide improved accuracy. Also, the paid tools have much better privacy and security features.  Even so, some free tools, including DeepL, are only free up to a point: after that, you’ll need a paid plan.  

To see why translation is worth investing in, let’s explore six risks of free machine translation that your business probably doesn’t want to mess with. 

Risk #1: Poor data security and privacy.   

To quote Jonathan Zittain, professor of internet law at Harvard Law School, “When something online is free, you’re not the customer, you’re the product.” That’s true here: in the case of free machine translation services, your data is the product. The content you send to be translated becomes training data to feed the AI language translation software’s ever-hungry algorithm.  

That means it’s no longer under your control. Imagine translating a sensitive internal memo, a high-stakes contract, or anything involving a consumer’s private data. When you use a free translation tool, you’re handing all of that over on a silver platter. If you’re using generative AI, like ChatGPT, there’s also the risk of confidential information or intellectual property being regurgitated in a future chat session.  

Risk #2: The output may not be accurate.  

If the tool is not customized, you don’t have any assurance that the output will meet your quality needs.  Machine translation accuracy depends on a number of factors, including the complexity of the source content, the language pairs being translated, and which MT engine you’re using. For example, a 2019 study found that Google Translate’s accuracy varied from 94% (English to Spanish) to 55% (English to Armenian).  
 
Translation errors can cause brand damage, negative publicity, and more. In high-stakes situations, the consequences can be devastating. For example, 40% of Afghan asylum cases encountered problems due to AI translation being used in the U.S. immigration system.  

While the initial translation may have been free and fast, fixing these mistakes probably won’t be.  

(Customization goes a long way in fixing these accuracy errors).  
Risk #3: The level of fluency may be disappointing. 

Translation quality is about more than just accuracy. Translation is not just converting words, it’s about translating ideas, cultural references and specific elements of style. Even if the end result is technically accurate, does it sound natural? Does it sound on-brand? Does it sound like you?  

If you used a free MT tool, it might not. On the other hand, with the right MT tools and a careful review from a skilled linguist, you can make sure your brand voice comes through loud and clear.  

Risk #4: You may be exposed to liability risks.  

Free MT tools come with liability risks due to their (lack of) data security practices and potential for inaccuracy. Imagine translating a contract, safety guidelines, or even user terms and conditions with a free machine translation tool. One mistranslated clause, one overlooked term, and you could find yourself facing costly legal repercussions.  

Without the accuracy and quality control that come with professional, paid translation services, you're rolling the dice with every document you translate. When contractual obligations, compliance issues, or even consumer rights are on the line, “We used a free translation tool” will not protect you from the liability that results.  

Risk #5: AI translation services can be biased.  

We’ve all heard about bias in AI. AI is not inherently unbiased - the cultural, racial, and gender biases come baked into the training data it learns from. For instance, Google Translate has been known to perpetuate gender stereotypes, defaulting to gendered language where it shouldn't.  

When these biases seep into your business communications, you run the risk of alienating your audience and perpetuating harmful stereotypes.  

Risk #6: Some AI tools can hallucinate. 

Generative AI tools like ChatGPT tend to hallucinate.  If the tool doesn’t know the answer to a question, it will make one up. If you’re using ChatGPT translation, the output may include words or phrases that are not present in the original text. For businesses, this could have unpleasant and expensive consequences. For example, imagine your AI tool "hallucinates" a term or a clause that alters the meaning of a legal document or an instruction manual.  

The real cost of machine translation 

If even one of these risks came to pass, implementing MT might not have been worth it for your business. Well, sadly, you do get what you pay (or don’t pay) for.  

When it comes to enterprise-level needs, machine translation is never really free. Given the downsides—from unpredictable accuracy and quality to inadequate security—the answer is usually no. It’s simply not worth the risk. 

You need secure, customized MT tools fine-tuned to take into account industry-specific terms and to preserve your brand voice. Before you can train an MT engine on your data, that data needs to be cleaned up to fix errors and inconsistencies (otherwise the MT engine may not draw the correct conclusions). You need to build out a workflow between the machine translation engine, translation memory (TM) tool, and any quality tools.  You also need skilled linguists on-tap to post-edit translations and catch machine-generated mistakes before they go live.  

Customizing an MT engine to fit your company’s specific needs takes time and expertise, and so no, it’s not free. Post-editing is not free, either. However, when deployed correctly, an MT program can massively streamline your translation workflow. A strategic MT program is an investment, and with productivity gains of up to 60% in some cases, it’s well worth it.   

We can help you design a strategy that will guide you to success, contact us for a consultation today!