It’s still all about Deepseek. The Financial Times reports that OpenAI has found evidence that Chinese Artificial Intelligence startup Deepseek used the US company’s proprietary models to train its own open source competitor. What recourse does AI have to counter this? Probably not much.
Many people have pointed out the irony of OpenAI complaining that someone has stolen their training data, asking “Where exactly did you get your training data OpenAI?” Meanwhile, Perplexity has announced that the DeepSeek R1 model is now available as part of their search experience. The limits are low for now, but they’re promising to increase it every day. That’s a very impressive turnaround time from Perplexity.
The Grok standalone app has got a new addition in its interface. The ‘Thinking Time’ clock hints at what might be to come from Grok 3. Should we be expecting it to be a brand new reasoning model?
Krea has announced that it’s added character consistency via the Hailuo model to its tools. There have been so many innovations in AI video models recently that it’s hard to keep up, but character consistency is absolutely crucial in this market.