August 2023

August 2023 focuses sharply on the gap between AI's practical deployment and the safety infrastructure needed to support it. A supermarket meal-planning chatbot generating a chlorine gas recipe serves as a vivid illustration, while Snyk's research into AI hallucinations producing software vulnerabilities and Google's red team outlining four common AI attack patterns signal that enterprise security teams are taking the threat seriously. On the policy front, Australia's Department of Industry opens a consultation on responsible AI, Digital Rights Watch submits to it, and Atlassian publicly calls for transparency disclosures when users interact with AI bots — a notable corporate position in the Australian context. The disinformation dimension is quantified bluntly: a $400 disinformation machine is now within reach, while Google's watermarking research for AI-generated images offers a nascent countermeasure. Shadow AI adoption runs parallel to all of this, with employees secretly using ChatGPT to advance their work regardless of corporate bans.