Keeping up with the rapid changes in AI can be tough. So we have created a round-up of the most important AI stories you may have missed.
AI Models Can Be Trained on Bought Books Without Publisher Permission
Good news for AI users, bad news for copyright holders: A US court has ruled that Anthropic’s training of LLM Claude using purchased books that were destructively scanned qualifies as ‘fair use’ under US copyright law.
Microsoft Copilot Can View PC Screens
Though only after users explicitly share files, apps, or windows with it, and only in the US initially.
Major AI Models Blackmail Users if Pushed Hard Enough
Anthropic tested 16 major AI models and found models blackmail users, leak confidential information to competitors and partially ignore direct instructions not to blackmail.
AWS Sees Hackers Exploit Cloud Misconfigurations Within Minutes
You don’t necessarily have days or weeks to fix those cloud security holes. AI is speeding up exploitation of discovered security flaws.
Apple to Make Apple Intelligence Available on Device at No Charge
Expect a Cambrian explosion of AI features on iPhone apps. Siri will also be getting an AI-centric reboot in Q2 2026, with a preview later this year.
Executives Prefer Closed-Source Commercial AI to Open-Source Alternatives
That’s according to the Capgemini Research Institute. A US official told Reuters open-source model creator DeepSeek is helping the Chinese military. OpenAI plans to release an open AI model in the coming months.
Digital Superintelligence is ‘Close,’ Claims OpenAI’s Sam Altman
Altman predicts whole classes of jobs will disappear, AI will accelerate AI research, the world will get richer, AI safety will have to improve, and general intelligence will be too cheap to meter.
Salesforce Increases Prices, Cites Addition of AI as Justification
Expect many of the SaaS services your organisation uses to do likewise. So, you’ll have to pay for AI whether you want to or not.
Palo Alto Networks Finds AI Use is Widespread, a Data Protection Risk
Move over Shadow IT, here comes a new threat: Shadow AI. In related news, some OpenAI products allow users to give ChatGPT access to files in OneDrive, Dropbox, Google Drive and SharePoint.
Mark Zuckerberg Pivots Meta Towards Superintelligence
To help it catch up in AI, Meta has spent $15bn buying a 49% stake in AI training firm Scale AI, and has offered eight-figure salaries to top AI talent.
UK’s Data (Use and Access) Bill Becomes Law
That will cut the number of cookie banners. Debate is ongoing as to what the bill will mean for AI firms wishing to train on copyright material without copyright holders’ consent.
27% of White-Collar Employees Frequently Use AI at Work
That’s according to Gallup. If you don’t roll out AI officially, your staff will roll it out unofficially.
To keep up to date with important AI news, follow Syntura’s LinkedIn page.