Thought #24 - Trust, Tension, and Too Many Releases

Hi lovely humans,

I did jinx myself by calling last week a slow one. Everyone’s clearly back from holidays and dropping new their releases on a random Tuesday (yesterday).

Some big ones too - OpenAI has released open source models, Anthropic has launched Claude Opus 4.1, and Google’s put out a few impressive tools of their own.

The rivalry between OpenAI and Anthropic also seems to have escalated, with Anthropic banning OpenAI devs from using Claude to build GPT-5.

Elsewhere, there are plenty of job market updates, plus some genuinely interesting use cases from companies and government - including a few that are saving serious money.

There were 35 news stories we included this week (not an all-time high, but definitely up there. Suggesting the current AI wave is far from slowing down.)

Our Week in AI

A Reminder About Trust

My LinkedIn was flooded earlier in the week about shared ChatGPT conversations being searchable on Google.

(if you specifically searched for “https://chatgpt.com/share,”, some chats were founds)

Rage and panic were the response from many as personal conversations were found.

Now I’m really for data privacy - but I’m actually on OpenAI’s side in this one.

It was very clear that there was no privacy settings for sharing these (like in Google or Microsoft where you restrict to specific people, or make the doc viewable to anyone with a link).

OpenAI also stated that shared chats may be discoverable before you clicked the link.

I searched my own shared chats - 4 total. Only 2 were in the indexing, and these were the ones I’ve linked publicly on my blogs.

Just a little reminder to always read what the company is telling you, and not just assume it works as expected.

AI New Releases

As the majority of these were released yesterday or this morning, we have not yet had time to play around (updates next week):

OpenAI

• OpenAI has released two open source models - very big news

• These OpenAI models are now available on AWS - the first open source models hosted there

Anthropic

• Claude Opus 4.1 released - much to developers’ delight - ahead of the rumoured OpenAI GPT-5. Claude has been erroring more than normal recently (which usually happens around a release) so no one is surprised.

• Claude Projects now have view and edit permissions for Team and Enterprise

• Claude Artifacts can now read your team files directly

• Claude now has a LinkedIn account

Google

• Google launches Genie 3 - a new video creation model (looks impressive, we're keen to play)

• Google has released a "new AI coding teammate" - Gemini CLI GitHub Actions

• Google has launched Game Arena - an open source platform for testing AI models

Others

• Eleven Labs release an AI model for generating music - and claim it's okay for commercial use

• Deep Cognito has released open source models

AI News

• Anthropic claims OpenAI are using Claude to train GPT-5 - and has revoked access; OpenAI says it’s all fine as Claude is now part of coding infrastructure

• Kaggle has been running a chess competition for LLMs - Grok 4 is currently winning, and the matches are being written up brilliantly by chess.com

• Ministry of Justice AI Action Plan - a three year plan for rolling out AI, including the goal of all workers getting a personal AI assistant by December (and a lovely mention of a project we’re working on with the MoJ!)

• Cisco and Hugging Face have partnered on malware scanning - this will help keep uploaded files (including models) safer, which is needed given Hugging Face hosts 1.9 million models

• OpenAI, Google, Anthropic added to the US federal vendor list

• Wells Fargo is rolling out AI across the company using Google tools

• Police in the UK are using AI to investigate grooming gangs - apparently saving more than £20m and 16k hours of time

• In Bristol, the health minister has visited and commended a GP surgery for using AI to speed things up

• Cambridge United (football club) are rolling out a platform to help manage player contracts - using Genie AI

• Sweden’s PM criticised for using AI in his role - he said he used it for a “second opinion” which has caused concern from some tech experts

• BP is aiming to bring costs down, and hasn't ruled out using AI to replace jobs - after 8,000 workers and contractors were already let go in January

• Two Chinese nationals accused of illegally shipping Nvidia AI chips to China

• European power plants getting a repurpose as data centres - driven by the AI boom

• Australian creatives calling for action - they want the government to step in against the unauthorised use of copyrighted materials to train AI models

• UK MP in Leeds creates an AI version of himself

• Lovable (an AI platform that helps build websites and apps from prompts) has raised a $200M Series A - they’re a team of 50 and this is one of the largest ever in Europe

• Experts warn AI therapy bots might make the mental health crisis worse

• Anthropic has published a framework for developing safe and trustworthy agents

• Github is calling for open source funding in Europe - they argue we need a European sovereign tech fund

Not Quite News, But Worth a Read (or Listen or Watch)

• In this week’s Diary of a CEO, an ex-Google exec has taken a very pessimistic view - AI is definitely not going to create jobs and all of them will be gone by 2037 (this take makes more sense if you know what he’s selling)

• Goldman Sachs economist expects 6-7% of jobs globally to be replaced - but tech is still hiring (just not juniors)

LinkedIn AI Poll

We wanted to know what people thought “training models with your data” meant. And we’re pleased to say most people understood it doesn’t mean your chat will end up as data for other people’s chat.

An interesting one as 20% of people thought others might see there data or had no idea - suggesting a real need for clearer data privacy policies

Vote in this week’s poll - please!

This week we want to know if data privacy concerns impacts how you use AI tools.

Final Thoughts

As always we hope this was helpful!

Feel free to share this with anyone who might find it useful.

Next week, we will have actually used some of the new tools and will have some opinions. Praying for a less newsworthy week.

Laura
Always learning