A week in Generative AI: Enterprise, Coding & Deliveries
News for the week ending 8th June 2025
This week has mostly been new features and model updates. We’ve had a lot of new enterprise focused features from OpenAI, a new coding product from Mistral, and an updated Gemini 2.5 Pro with improved coding capabilities. Amazon are also rumoured to be starting to test humanoid robots to help assist delivery drivers.
Lots of Ethics News this week too. OpenAI have opened up about how they’re responding to the New York Times’ lawsuit and there’s lots of reports on copyright, as the debates rumble on, and Reddit is suing Anthropic for unauthorised use of their data.
In Long Reads I highly recommend Simon Willison’s post about how the UK technology secretary has been using ChatGPT. There’s also a great article from Joanne Jang, of OpenAI, on how they’re thinking about the human-AI relationship.
ChatGPT introduces meeting recording and connectors for Google Drive, Box, and more
This week OpenAI added lots of business-focused features to ChatGPT, mostly focused on integrations with popular enterprise platforms. These include connections for DropBox, Box, SharePoint, OneDrive, and Google Drive which covers most enterprise storage solutions.
ChatGPT can also now record and transcribe meetings. It doesn’t do this by integrating with any apps, you can just click the record button in the Mac App and it will transcribe any conversation it can hear via the Mac’s microphone. I’m not sure how useful this is or how well it works as the best AI models for meetings are fully integrated so they can understand who’s speaking when.
The most exciting part of these announcements though is the ability to use ChatGPT’s Deep Research feature across any of the enterprise integration it now has. This is a big game changer as I’ve long wanted Deep Research to have access to my own files alongside the research it can do on the web for me. This is now similar to how Anthropic’s Deep Research product works, although OpenAI have many more connectors now available.
Mistral releases a vibe coding client, Mistral Code
Nice to see the French company Mistral getting in on the coding agent bandwagon. This has become the hottest battleground between AI companies right not as not only is coding a great low-risk area to develop agentic capabilities but if it works it has a flywheel effect where it will speed up future development too.
This isn’t all Mistral’s work (it’s a version of the open source Continue coding project) but it does use lots of different Mistral models under the hood (Codestral, Devstral etc.) and also supports coding in 80 different languages.
Google says its updated Gemini 2.5 Pro AI model is better at coding
Speaking of coding… Google DeepMind have done another stealthy update of Gemini 2.5 Pro which it claims is now better at coding. This is likely a response to the recent progress both OpenAI and Anthropic have been making with their models for coding and features like Codex/Claude Code.
The new model will be generally available in “a couple of weeks” and alongside coding improvements, it also performs very highly on maths, science, knowledge, and reasoning benchmarks.
Amazon to test humanoid robots for package delivery with Rivian electric vans
Interesting…. I’m guessing that this testing is getting ready for when the vans can drive themselves so deliveries can be made without any humans involved. If this does end up working and the tests go well, then this is probably the first exposure most people will have to a humanoid robot. I would never have guessed that my first experience would be a delivery!
AI Ethics News
OpenAI: How we’re responding to The New York Times’ data demands in order to protect user privacy
Alphabet CEO Sundar Pichai dismisses AI job fears, emphasizes expansion plans
Government AI copyright plan suffers fourth House of Lords defeat
BBC and Sky bosses criticise plans to let AI firms use copyrighted material
Reddit sues Anthropic for allegedly not paying for training data
Lawyers could face ‘severe’ penalties for fake AI-generated citations, UK court warns
Anthropic unveils custom AI models for US national security customers
AI pioneer announces non-profit to develop ‘honest’ artificial intelligence
Anthropic’s AI is writing its own blog — with human oversight
Long Reads
Simon Willison - Tips on prompting ChatGPT for UK technology secretary Peter Kyle
One Useful Thing - The recent history of AI in 32 otters
FT - ‘Humanity deserves better’: Jony Ive and Laurene Powell Jobs on tech’s next chapter
Joanne Jang - Some thoughts on human-ai relationships and how we're approaching them at OpenAI
“The future is already here, it’s just not evenly distributed.“
William Gibson