Mid-September put AI assistants directly into the software people already use. Apple shipped its rebuilt Siri in beta and Salesforce arrived inside Claude. The same week, a European regulator logged what it calls the first data breach carried out by an AI agent, the House voted on who pays for AI's electricity, and a global survey showed how worried workers are.
Apple ships Siri AI in public beta
Apple released Siri AI in beta in English on September 14 across iPhone, iPad, Mac, Apple Watch and Vision Pro, according to Apple's announcement. Apple says the models behind it were custom-built in collaboration with Google and its Gemini models, and run on the device and on Apple's Private Cloud Compute servers. Siri AI can take actions inside some third-party apps, with Microsoft Outlook support coming soon. It is not initially available on iPhone in the EU, and server features carry daily limits, with expanded access to be offered for a fee later.
Our take: Most of your staff carry an iPhone, so this is the AI assistant many of them will use first. Decide whether work email and client files on personal phones should be reachable by it, and say so in writing.
Salesforce lands inside Claude
Anthropic released Salesforce in Claude, a beta plugin with 37 skills for sales work such as account research, call prep, pipeline review and CRM updates, in a September 15 post. Sellers sign in with their Salesforce credentials, Claude sees only what their permissions allow, and by default every change needs the seller's approval before it is written back. Anthropic says about 7,000 Salesforce sellers already use it.
Our take: This is the pattern to look for in any AI tool you buy: it works inside your existing system, respects existing permissions and asks before it changes records. If a vendor cannot explain those three things, keep looking.
Spain logs a data breach carried out by an AI agent
Spain's data protection agency described what it calls the first notified personal data breach carried out through an AI agent, SecurityWeek reported on September 16. According to the report, the agent logged in to a Spanish firm's systems, searched for weaknesses and gained the ability to modify personal data and access invoices. The cause is not confirmed, and the investigation is continuing.
Our take: Regulators now treat agent-driven breaches as a real category, which means your breach response plan should too. Check that logins to your business systems use multi-factor authentication, since an agent that can log in can do whatever that account can do.
House votes to make data centers pay for their power upgrades
The House passed the Ratepayers Protection Act 417 to 3 on September 16, NBC News reported. The bipartisan bill would set a federal standard that state utility regulators must consider, aimed at making large data centers, not household customers, cover the cost of new generation and transmission built to serve them. Senate action is uncertain, and NBC says further movement is not expected until after the midterms.
Our take: AI's power bill is turning into a pocketbook issue. Whether or not this bill becomes law, expect your utility rates and your cloud AI prices to keep reflecting the cost of new data centers.
Most of the world expects AI to cost jobs
People in 34 of 37 places surveyed think AI will lead to fewer jobs rather than more, according to a Pew Research Center report released September 17. In the US, about seven in ten adults expect AI to cause job loss over the next 20 years, a share Pew says rose 7 points in two years. The survey measures expectations, not actual job losses.
Our take: Your employees are reading the same headlines. When you bring in an AI tool, tell people what it is for, what it is not for, and how it changes their work, before the rumor mill does it for you.
One thing to do this week
Make a one-page list of every AI assistant your team can reach from work devices, including Siri AI, Gemini, Copilot and ChatGPT. Next to each, write whether it may touch client data. Share it with the team.
Want a system like this in your business?
A discovery conversation is where we map it: what to automate first, where the human checkpoints go, and what it would take.
Book a discovery conversation →