19th August, 2026

AI guardrails and governance, anyone?

THIS ARTICLE breaks down what we know about recent AI “sandbox breaches”, and why we should be advocating for AI governance/legislation in our respective countries. Closer to home, we should also consider whether our workplaces have established AI governance. In Shelagh’s past annual GenAI surveys, a significant percentage of respondents reported there was either no communication on the matter or that policy development was underway. Check back here August 20, 2026 for Shelagh’s latest annual AI survey, and take a couple of minutes to add your voice and experiences to the conversation. You’ll find 12 multiple choice questions you can answer in the blink of an eye.

What should you know about an unsettling matter Anthropic, Meta/Facebook and OpenAI share in common in 2026?

It’s less than a month since Anthropic and OpenAI reported incidents in which their advanced AI models went rogue while using software from a startup called Irregular. In some simulations, OpenAI’s agents independently hacked servers at another startup, Hugging Faces. This led to Irregular auditing its systems, and the resulting news that both Anthropic’s AND Meta’s/Facebook’s systems had also been functioning in similar ways.

On August 5, 2026, the Harvard Gazette stated that the UK’s AI Security Institute had “… reported additional incidents, saying AI agents created fake online personas in order to improperly access real people and companies during security tests it conducted on the two tech firms.” On August 7, 2026, Forbes suggested the OpenAI breach was even more alarming than initially thought. How so? We’re informed that, after OpenAI dismantled the first network, its AI agents simply used another method to rebuild it.

We’re told the AI agents’ software activities took place in the sandbox.

“Sandbox” is the term for a “secure”, isolated testing environment designed to explore AI’s capabilities while mitigating a number of privacy and security risks. That doesn’t make the news any less unsettling, especially at a point in time when workplace decision makers and leaders are either grappling with how – or whether – to best make use of GenAI, and how to govern AI usage.

On that note, while the European Union has established the EU AI Act, and Denmark has gone a step further in seeking to protect individuals’ privacy and protection against deepfakes. Other countries around this planet need to step up in terms of AI governance.

We need workplace AI governance and guardrails.

Has your employer established GenAI or agentic AI policies or guidelines? In 2023, I began conducting surveys/Weekend Polls so we can see the extent to which our work environments and even careers have changed since the emergence of ChatGPT in late November 2022. To date, it’s been striking to see the percentage of respondents who said policy development was either (a) still (the “still” reference is my own, and not respondents’) underway, or that (b) there’d been no communication from their employers regarding workplace use of generative AI (GenAI).

I’ll be publishing this year’s AI survey on Thursday, August 20, 2026. It includes 12 multiple choice questions, so it should take almost no time for you to add your voice and anonymously document your experience (or lack of) with GenAI. As AI continues to make inroads in our world, I’ve switched up some of the questions from previous years, and look forward to sharing the results with you.

I’.ve been writing about AI and what I used to refer to digital disruption for years now, including some interviews and articles that now seem rather prescient. I’ve bundled links to some of these pieces on the “AI – Artificial Intelligence” option you’ll see atop your screen. Have a look at my AI page if you’d like to see how far we’ve come in the space of less than a decade.