The internet is full of translation tools, capable of converting text between countless languages. But what if you want to “translate” into something a little less conventional – like the jargon-heavy world of “LinkedIn Speak,” the rapidly evolving slang of Gen Z, or even the imagined pronouncements of a historical figure? This week, users discovered that Kagi Translate, an AI-powered translation service, can attempt just that, highlighting both the creative potential and the inherent risks of increasingly powerful large language models.
The ability to generate text in these unconventional “languages” quickly spread online, prompting both amusement and discussion about the boundaries of AI and the nature of language itself. While Kagi Translate’s playful functionality demonstrates the flexibility of modern AI, it also raises questions about responsible use and the potential for misuse of these tools. The core of this story is Kagi Translate’s AI answers the question “What would horny Margaret Thatcher say?” and the broader implications of AI-powered translation beyond traditional language barriers.
Kagi Translate: Beyond Traditional Translation
Kagi, launched in 2024, initially positioned itself as a premium alternative to Google Search, addressing concerns about declining search quality. As Ars Technica reported in August 2025, many users were frustrated with Google’s search results and sought a more focused experience. The company then introduced Kagi Translate, aiming to surpass established players like Google Translate, and DeepL. At its launch, Kagi emphasized that its translation tool utilized a combination of large language models (LLMs) to deliver superior results, acknowledging that this approach “can occasionally lead to quirks that we’re actively working to resolve.”
Initially, Kagi Translate offered translation between 244 languages via standard dropdown menus. Still, as early as February 2025, users began discovering hidden functionalities. A poster on Hacker News noted the ability to manipulate URL parameters to specify unusual target languages, such as “rude man with a Boston accent,” without causing errors. The Hacker News post demonstrated that the underlying AI was surprisingly adaptable to unconventional requests.
From LinkedIn Speak to… Margaret Thatcher?
More recently, Kagi itself began showcasing the tool’s ability to mimic specific styles of communication. The company’s social media accounts highlighted its capacity to generate text in “Reddit Speak” or “McKinsey consultant speak.” Kagi demonstrated the “Reddit Speak” translation on Bluesky, and showcased “McKinsey consultant speak” on the same platform. However, the functionality truly gained widespread attention when a Hacker News user reported that Kagi Translate now supported “LinkedIn Speak” as an output language. Further exploration revealed that users could directly type their desired “language” into the search bar, and the AI would attempt to accommodate the request.
This led to a wave of experimentation, with users prompting the tool to generate text in increasingly bizarre and provocative styles. One particularly notable example, reported by Ars Technica, involved asking Kagi Translate to respond as if it were Margaret Thatcher, with…unconventional inclinations. The results, while not explicitly detailed, sparked a conversation about the ethical implications of allowing users to manipulate AI in this way. As reported by Google News, Kagi Translate’s response to this prompt raised eyebrows and prompted discussion about the boundaries of AI-generated content.
What Does This Mean for AI and Language?
The Kagi Translate experiment underscores a fundamental question: what *is* a language? Traditionally, language is defined by grammar, vocabulary, and shared cultural understanding. But Kagi Translate demonstrates that AI can be prompted to mimic stylistic patterns, jargon, and even the perceived personality of an individual, effectively creating a new form of “language” based on data and algorithms. This raises concerns about the potential for misuse, including the creation of convincing but fabricated content or the amplification of harmful stereotypes.
The incident also highlights the challenges of controlling the output of large language models. While developers can implement safeguards, users are often able to find creative ways to circumvent these restrictions. The open-ended nature of AI, while enabling innovation, also presents risks that demand to be carefully considered. The ability to generate text in “LinkedIn Speak” or mimic the voice of a public figure demonstrates the power – and potential pitfalls – of this technology.
Kagi has not yet issued a formal statement regarding the recent discoveries, but the company’s initial response suggests an awareness of the issue and a commitment to addressing potential concerns. As AI technology continues to evolve, developers will need to grapple with the ethical and practical challenges of balancing innovation with responsible use. The next step for Kagi will likely involve refining its algorithms and implementing more robust safeguards to prevent the generation of inappropriate or harmful content.
What are your thoughts on AI-powered translation and the potential for unconventional “languages”? Share your comments below and let us understand how you think this technology will evolve.
