AI Weekly: The Tool Everyone Uses Gets More Reliable — May 4-10, 2026

The most-used AI tool in the world got quietly more accurate this week, and Europe eased off the timeline for its AI rules. Neither was a flashy product launch, but both matter for everyday business. When the default ChatGPT model hallucinates less and lands native spreadsheet support, that reaches far more small businesses than any benchmark.

The Deep Dive: The Tool Everyone Uses Gets More Reliable

GPT-5.5 Instant becomes the ChatGPT default

Around May 5, OpenAI made GPT-5.5 Instant the default model for all ChatGPT users, citing roughly 52.5% fewer hallucinations on high-stakes prompts. ChatGPT for Excel and Google Sheets also reached general availability. Most people never change the default model, so making the default better instantly improves the experience for millions who will never touch a settings menu.

The spreadsheet support is the practical win. A lot of small-business work lives in a spreadsheet, and being able to ask questions of your own data without exporting it anywhere removes a real friction point. For most teams, that is a bigger day-to-day change than any new frontier model.

Europe eases the AI Act timeline

On May 7, the EU agreed a "Digital Omnibus" deal that defers the high-risk obligations of the AI Act from August 2026 to December 2027, while pushing synthetic-content labelling to December 2026 and adding new prohibitions on abusive AI-generated imagery. More breathing room to comply, but labelling rules are still coming.

Why this matters:

• A more accurate default model means better results for the many staff who use ChatGPT as-is.
• Native spreadsheet support brings AI to where a lot of SME work actually happens.
• Even outside the EU, the AI Act shapes how global vendors build, so its rules ripple to Australian users too.

What This Means for Australian SMEs

The accuracy improvement is a free upgrade you do not have to install. If your team already uses ChatGPT for drafting, summarising or answering questions, it just got more trustworthy. Still check its work on anything that matters, but the baseline is better.

The spreadsheet feature is worth a real trial. If you spend time wrangling numbers, being able to ask plain-English questions of a sheet, like "which customers ordered less this quarter", can save hours. On the regulation front, the labelling rules are the part to note. If you publish AI-generated images or video, expect "this was made with AI" disclosure to become standard practice, so it is worth getting into the habit now.

The practical takeaway for this week:

• Retest a task where ChatGPT used to get things wrong. The new default may handle it better.
• Trial ChatGPT inside Excel or Sheets on one real reporting job this week.
• If you publish AI-generated media, start labelling it now. It is heading toward being expected everywhere.

Helping SMEs use these everyday tools well, and stay on the right side of the rules, is part of what we do at IntelliAgent. If you want a hand getting more out of the AI tools you already pay for, get in touch.