Google seals AI training deal with Reddit

Google and social media platform Reddit have formed a partnership where Reddit will supply its content for training the search engine giant’s artificial intelligence (AI) models. In an announcement, Reddit said it will offer Google improved methods for training models.

Through the collaboration , Google can use Reddit’s data application programming interface (API), which provides real-time content from Reddit’s platform, allowing access to Reddit’s extensive content and enabling the display of Reddit content across Google’s products.

Reddit will access Vertex AI, Google’s AI-powered service designed to enhance search results for companies. Reddit clarifies that this update doesn’t impact the terms of its data API, maintaining restrictions on commercial access without approval for developers or companies.

This partnership with Google comes after Bloomberg reported that Reddit secured a $60 million training agreement with an undisclosed AI company. Reddit announced a plan to charge companies for API use in the past year. The partnership with Google marks Reddit’s first known agreement with a major company building AI.

In 2023, Google updated its privacy policy to allow it to use publicly available data for AI training. This update came shortly after ChatGPT developer OpenAI was sued in  a class-action lawsuit in California over allegedly scraping private information from users via the internet.

Related: Google updates Gemini AI, apologizes for ‘woke’ inaccurate imagery

However, according to updates to the Claude developer’s commercial terms of service, generative AI startup Anthropic promised not to use client data for large language model (LLM) training from January 2024.

Despite the deal, Google and Reddit haven’t always seen eye to eye. Reddit previously threatened to block Google from using its crawlers on its site over concerns that companies would use its data for free to train AI models.

After a yearslong run-up , Reddit filed its initial public offering on Feb. 22 to boost its valuation, which sat at more than $10 billion in 2021. The IPO filing, expected to go public in March, will be the first major social media IPO since Pinterest’s in 2019.

Makers of AI models have been actively securing agreements with content owners in recent months to broaden their training data beyond extensive web scraping, as numerous content owners claim their material was utilized without permission .

Magazine: Google to fix diversity-borked Gemini AI, ChatGPT goes insane: AI Eye