# Daniel De Freitas

> Source: https://aiwiki.ai/wiki/daniel_de_freitas
> Updated: 2026-11-18
> Fact-checked: 2026-11-18
> Categories: Conversational AI, Google, People
> License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) - attribute to "AI Wiki (aiwiki.ai)"
> Cite as: AI Wiki. "Daniel De Freitas." aiwiki.ai, 18 Nov 2026. https://aiwiki.ai/wiki/daniel_de_freitas
> From AI Wiki (https://aiwiki.ai), the free encyclopedia of artificial intelligence. Reuse freely with attribution.

**Daniel De Freitas** is a Brazilian engineer who started the Google chatbot project that became Meena and later [LaMDA](https://aiwiki.ai/wiki/lamda), and who co-founded the chatbot company [Character.AI](https://aiwiki.ai/wiki/character_ai) with [Noam Shazeer](https://aiwiki.ai/wiki/noam_shazeer) in 2021 [1][2]. After studying computer engineering at the University of São Paulo and working at [Microsoft](https://aiwiki.ai/wiki/microsoft) from 2012, he joined [Google](https://aiwiki.ai/wiki/google) in 2016, where he began the chatbot work as a side project and led its development [1][2][5]. He and Shazeer left Google in late 2021 after executives repeatedly declined to release the chatbot publicly, according to reporting by The Wall Street Journal [6]. In August 2024 he returned to Google, together with Shazeer and other Character.AI staff, as part of an agreement under which Google took a non-exclusive license to Character.AI's technology [12]. He was later named as an individual defendant in Garcia v. Character Technologies, a wrongful death suit over a teenager's suicide that was dismissed after the parties settled in January 2026 [14][16].

## Early life and education

De Freitas grew up in Brazil [3]. Estadão describes him as a native of São Paulo and reports that he graduated in computer engineering from the University of São Paulo (USP), spent time at [Stanford University](https://aiwiki.ai/wiki/stanford_university), and then took a job with Microsoft in the United States rather than pursuing an academic career in Brazil [2]. His interest in chatbots started in childhood: a 2022 Washington Post profile says he wrote his first chatbot at the age of nine [3], while a 2023 post by the venture firm [Andreessen Horowitz](https://aiwiki.ai/wiki/andreessen_horowitz) put the age at seven [4]. Shazeer has described De Freitas as having been on a "lifelong mission" to build intelligent chatbots [5].

His full name includes the surname Adiwardana: a 2022 Washington Post photo caption calls him Daniel de Freitas Adiwardana [3], and the Meena paper and Google's announcement of it list their first author as Daniel Adiwardana [7][8]. Court papers in Garcia v. Character Technologies also give the additional surname, which the filings spell inconsistently (the January 2026 order's caption reads "Daniel De Freitas Adiwarsana") [14][16].

<figure class="embedvideo">
<iframe src="https://www.youtube-nocookie.com/embed/roVOIW7ho58" title="Google's Meena: Open Dialog systems | Daniel De Freitas Adiwardana"></iframe>
<figcaption>"Google's Meena: Open Dialog systems | Daniel De Freitas Adiwardana", video from Rasa, published July 7, 2020. <a href="https://www.youtube.com/watch?v=roVOIW7ho58">Watch on YouTube</a>.</figcaption>
</figure>

## Career

| Date | Event |
|---|---|
| 2012 | Joins Microsoft as a software engineer in the United States [2] |
| 2016 | Joins Google [5] |
| 2020-01 | Google announces the Meena chatbot [8] |
| 2021-05 | Google announces LaMDA [9] |
| 2021 (late) | Leaves Google with Noam Shazeer [6] |
| 2021-11 | Co-founds Character.AI [2] |
| 2022-01 | Second author of the LaMDA research paper [10] |
| 2022-09 | Character.AI opens a public beta [3] |
| 2024-08 | Rejoins Google under the Google and Character.AI licensing agreement [12] |
| 2024-10 | Named as a defendant in Garcia v. Character Technologies [14] |
| 2026-01 | Garcia case dismissed after the parties report a settlement [16] |

### Microsoft (2012-2016)

De Freitas began working at Microsoft in 2012 as a software engineer in one of the company's U.S. offices [2]. According to Estadão, in his final year there he worked on natural language models and AI for Bing, Microsoft's search engine, and after about four years he was approached by a Google recruiter [2].

### Google and Meena (2016-2020)

Shazeer said in 2023 that De Freitas joined Google in 2016 after reading some of its research papers on language technology, and that he saw a chance to use the company's language research to build a chatbot [5]. Andreessen Horowitz's account places him first at [YouTube](https://aiwiki.ai/wiki/youtube) and later at [Google Brain](https://aiwiki.ai/wiki/google_brain), and says he created Meena during that period [4]. Business Insider, summarizing The Wall Street Journal's reporting, wrote that he started the chatbot as an AI side project around 2018, while he was a research engineer at Google [6].

The project had few dedicated staff. Shazeer told the No Priors podcast: "He did not get a lot of headcount. He started the thing as a 20% project," referring to Google's practice of letting employees spend part of their time on side projects, and added that De Freitas "recruited an army of 20% helpers who were ignoring their day jobs" [5]. Shazeer later joined the effort [6].

Google announced the result, Meena, in January 2020 [8]. The paper describing it, "Towards a Human-like Open-Domain Chatbot", presented Meena as a 2.6-billion-parameter multi-turn open-domain chatbot trained end to end, on data mined and filtered from public domain social media conversations, simply to minimize the [perplexity](https://aiwiki.ai/wiki/perplexity) of the next token [7]. The paper's first author is listed as Daniel Adiwardana [7]. Google's announcement post described the model as an Evolved Transformer sequence-to-sequence network with one encoder block and 13 decoder blocks, trained on 341 GB of filtered text, with 1.7 times the model capacity of [GPT-2](https://aiwiki.ai/wiki/gpt-2) and 8.5 times as much training data [8].

The Meena paper also proposed a human evaluation metric, Sensibleness and Specificity Average (SSA), and reported a strong correlation between perplexity and SSA [7]. The full version of Meena, with a filtering mechanism and tuned decoding, scored 79% SSA, which the authors said was 23 points higher in absolute SSA than the existing chatbots they evaluated; they estimated average human performance at 86% [7][8]. Google said it was not releasing an external research demo of Meena at that point, citing safety and bias concerns [8].

### LaMDA and the dispute over release (2020-2021)

Meena was later renamed LaMDA [5][6]. Google introduced LaMDA publicly on May 18, 2021, as a language model trained on dialogue that could hold open-ended conversations on many topics [9]. The full research paper, "LaMDA: Language Models for Dialog Applications", appeared in January 2022 with Romal Thoppilan as first author and De Freitas as second author, followed by Jamie Hall and Noam Shazeer [10]. It described a family of [Transformer](https://aiwiki.ai/wiki/transformers)-based dialogue models with up to 137 billion parameters, pre-trained on 1.56 trillion words of public dialog data and web text, and reported that fine-tuning with annotated data and letting the model consult external knowledge sources improved safety and factual grounding [10]. An essay introducing Character.AI, published in December 2022 and signed by both founders, says De Freitas "invented and led the development of Meena and LaMDA" [1]. LaMDA later powered [Bard](https://aiwiki.ai/wiki/bard), the experimental conversational AI service Google announced on February 6, 2023 [11].

According to The Wall Street Journal's reporting, as summarized by Business Insider, Google executives told De Freitas and Shazeer that the chatbot did not meet the company's standards for AI safety and fairness, and blocked their attempts to send it to outside researchers, add it to [Google Assistant](https://aiwiki.ai/wiki/google_assistant), or release a public demo [6]. The two left Google near the end of 2021 to start their own company, even though CEO [Sundar Pichai](https://aiwiki.ai/wiki/sundar_pichai) personally asked them to stay and keep working on the chatbot [6]. Shazeer later said Google had not believed a chatbot would gain enough traction to justify the reputational risk [5]. De Freitas and Shazeer told The Washington Post in 2022 that they had left Google to get the technology into as many hands as possible [3].

## Character.AI (2021-2024)

De Freitas and Shazeer founded Character.AI in November 2021 [2]. It opened a public beta in September 2022, letting users chat with characters modelled on public figures and fictional personas [2][3]. Shazeer served as chief executive [12], and De Freitas was the company's president [5]. In an interview with The Washington Post shortly after launch, De Freitas said, "We're trying to educate people as well," adding, "We have that role because we're sort of introducing this to the world" [3]. The same report said that his mother was learning English and that he had encouraged her to practise with Character.AI [3].

The founders' December 2022 essay introducing the company was signed "Noam Shazeer, Daniel de Freitas and the Character team" [1]. In March 2023 Andreessen Horowitz announced it was investing in the company [4]. By August 2024 Character.AI had raised more than $150 million, mostly from Andreessen Horowitz, according to TechCrunch [12].

## Return to Google (2024)

On August 2, 2024, Character.AI and Google announced that Google would sign a non-exclusive agreement to use Character.AI's technology, and that Shazeer, De Freitas and some other Character.AI employees would join Google [12]. Google said Shazeer would join the [Google DeepMind](https://aiwiki.ai/wiki/google_deepmind) research team but did not give the exact roles of either founder [12]. By January 2026 Fortune described De Freitas as a research scientist at Google DeepMind [17]. Dominic Perella, Character.AI's general counsel, became interim chief executive, and the company said most of its staff would stay [12].

The agreement was reported to be worth $2.7 billion [13]. In May 2025 Bloomberg reported that the U.S. Department of Justice was examining whether the deal had been structured to avoid regulatory scrutiny, in what it described as an early-stage inquiry; Google had not been accused of wrongdoing and said it was happy to answer regulators' questions [13].

## Litigation

On October 22, 2024, Megan Garcia filed a wrongful death and product liability complaint in the U.S. District Court for the Middle District of Florida over the death of her 14-year-old son, naming Character Technologies, Shazeer, De Freitas, Google and Alphabet as defendants [14]. The complaint alleged that, before co-founding Character.AI, De Freitas led the LaMDA development team at Google, and that he and Shazeer had tried to convince Google to integrate their chatbot into Google Assistant [14]. See [Garcia v. Character Technologies](https://aiwiki.ai/wiki/garcia_v_character_ai) for the case in full.

In an order filed on May 21, 2025, Judge Anne C. Conway dismissed the claims against Alphabet without prejudice at the plaintiff's request, dismissed the intentional infliction of emotional distress claim, and denied the remaining motions to dismiss [15]. She also denied Shazeer's and De Freitas's motions to dismiss for lack of personal jurisdiction, allowing them to refile after the plaintiff had taken jurisdictional discovery [15]. On January 7, 2026, after counsel told the court the case had been settled, the court dismissed it [16]. Fortune reported on January 8, 2026 that Google and Character.AI had reached a settlement in principle in several lawsuits brought by families, including Garcia's, with families from states including Colorado, Texas and New York; the terms were not disclosed [17].

## See also

- [Noam Shazeer](https://aiwiki.ai/wiki/noam_shazeer)
- [Character.AI](https://aiwiki.ai/wiki/character_ai)
- [LaMDA](https://aiwiki.ai/wiki/lamda)
- [Bard](https://aiwiki.ai/wiki/bard)
- [Garcia v. Character Technologies](https://aiwiki.ai/wiki/garcia_v_character_ai)

## References

1. Shazeer, Noam; de Freitas, Daniel; and the Character team. "Introducing Character". Character.AI blog, December 5, 2022. https://blog.character.ai/introducing-character/
2. Arimathea, Bruna. "Daniel de Freitas: conheça primeiro bilionário da IA no País que vendeu empresa para o Google". Estadão, August 7, 2024. https://www.estadao.com.br/link/cultura-digital/daniel-de-freitas-conheca-primeiro-bilionario-da-ia-no-pais-que-vendeu-empresa-para-o-google/
3. Tiku, Nitasha. "'Chat' with Musk, Trump or Xi: Ex-Googlers want to give the public AI". The Washington Post, republished by The Spokesman-Review, October 9, 2022. https://spokesman.com/stories/2022/oct/09/chat-with-musk-trump-or-xi-ex-googlers-want-to-giv
4. Wang, Sarah. "Investing in Character.AI". Andreessen Horowitz, March 23, 2023. https://a16z.com/2023/03/23/investing-in-character-ai/
5. Maxwell, Thomas. "A former Google researcher behind a seminal AI paper describes how the company lost a top chatbot visionary". Business Insider, April 20, 2023. https://www.businessinsider.com/google-ai-characterai-ceo-noam-shazeer-chatbot-2023-4
6. Mok, Aaron. "Two Google engineers built a ChatGPT-like AI chatbot years ago, but execs reportedly shut it down due to safety concerns". Business Insider, via Yahoo Finance, March 9, 2023. https://finance.yahoo.com/news/two-google-engineers-built-chatgpt-095900839.html
7. Adiwardana, Daniel; Luong, Minh-Thang; So, David R.; Hall, Jamie; Fiedel, Noah; Thoppilan, Romal; Yang, Zi; Kulshreshtha, Apoorv; Nemade, Gaurav; Lu, Yifeng; Le, Quoc V. "Towards a Human-like Open-Domain Chatbot". arXiv:2001.09977, January 27, 2020. https://arxiv.org/abs/2001.09977
8. Adiwardana, Daniel; Luong, Thang. "Towards a Conversational Agent that Can Chat About...Anything". Google Research blog, January 28, 2020. https://research.google/blog/towards-a-conversational-agent-that-can-chat-aboutanything/
9. Collins, Eli; Ghahramani, Zoubin. "LaMDA: our breakthrough conversation technology". Google blog, May 18, 2021. https://blog.google/technology/ai/lamda/
10. Thoppilan, Romal; De Freitas, Daniel; Hall, Jamie; Shazeer, Noam; et al. "LaMDA: Language Models for Dialog Applications". arXiv:2201.08239, January 20, 2022. https://arxiv.org/abs/2201.08239
11. Pichai, Sundar. "An important next step on our AI journey". Google blog, February 6, 2023. https://blog.google/technology/ai/bard-google-ai-search-updates/
12. Mehta, Ivan. "Character.AI CEO Noam Shazeer returns to Google". TechCrunch, August 2, 2024. https://techcrunch.com/2024/08/02/character-ai-ceo-noam-shazeer-returns-to-google/
13. Kabir, Omer. "Google's $2.7B AI deal with Noam Shazeer's Character.AI draws DOJ attention". Ctech by Calcalist, May 25, 2025. https://www.calcalistech.com/ctechnews/article/sy06wllflg
14. U.S. District Court for the Middle District of Florida. "Complaint, Garcia v. Character Technologies, Inc., No. 6:24-cv-01903 (Doc. 1)". Filed October 22, 2024. https://storage.courtlistener.com/recap/gov.uscourts.flmd.433581/gov.uscourts.flmd.433581.1.0.pdf
15. Conway, Anne C. "Order on motions to dismiss, Garcia v. Character Technologies, Inc., No. 6:24-cv-1903 (Doc. 115)". U.S. District Court for the Middle District of Florida, filed May 21, 2025. https://storage.courtlistener.com/recap/gov.uscourts.flmd.433581/gov.uscourts.flmd.433581.115.0.pdf
16. Conway, Anne C. "Order, Garcia v. Character Technologies, Inc., No. 6:24-cv-1903 (Doc. 244)". U.S. District Court for the Middle District of Florida, January 7, 2026. https://storage.courtlistener.com/recap/gov.uscourts.flmd.433581/gov.uscourts.flmd.433581.244.0.pdf
17. Nolan, Beatrice. "Google and Character.AI agree to settle lawsuits over teen suicides linked to AI chatbots". Fortune, January 8, 2026. https://fortune.com/2026/01/08/google-character-ai-settle-lawsuits-teenage-child-suicides-chatbots/

