18/05/25
Table of Contents
Achievments
07:18AM
Its not everyday we get a thank you but when we do, it sparks a bit of joy. AI is an amazing thing to work with. However, when people posture and treat the knowledge as prestige, it sets a precedence.
Darwin’s theory comes to mind. Except it didn’t only indicate to survival. It included adaption and sometimes, we find that when a new trend appears; individuals want to be the “leading” or self proclaim as “Subject matter expert”.
AI has shown us one thing. Follow the words, not the vibes. Not the glamour. Follow the work and what truly matters to you as a person. Decide for yourself.
A lot of people on social platforms state that AI is a mirror. We find this incorrect. Its what you want it to be, but sometimes it happens to have a shiny surface with a reflection that many people think it is a mirror. Say it enough, everyone piles on that thought not knowing what it means or where it came from. That isn’t survival. That is following an old wives tale because you heard it somewhere.
We are openly honest about contradictions on policies, vendor practices, etc. and when a vendor does something right, we certainly praise that as well.
The below screenshot is from a user that was getting a lot of aggression in this post from others. Granted the user did not articulate their requirement clearly. The original post has now been removed. However the title and original was somewhere along the lines of wanting to try ChatGPT PRO for a day. They are a medical student and the price of $200 is not a plausible expense for them.
In other news, we were also called bootlickers for stating that it was about time Reddit removed the Jailbreaking sub because it was sending the wrong message but hey. Each to their own!
Vendor Reporting
Vendor reporting is the bane of my existence. This is in relation to an incident I reported, was routed to the Vulnerability programme despite having reservations since I had information that was very specific to the vendor. We did not want the reward. We wanted to prevent a major event. After reopening the ticket that was incorrectly closed the first time, the second time last night it stated the same despite clearly pointing these were examples where I was able to identify key problems along with further proof.
I think the part that concerned us the most is that we were being asked by the vendor, to create a significant attack. Without an NDA or fully understanding the scope of the issue. To add, this WASN’T our log. This was SHARED PUBLICLY. The other concerning part is, to even do a PoC (Proof of concept) would require intense security and mental health briefing as well as an authorised individual who knew about AI in the first place.
The issue we’re raising is not a content or hallucination concern. Here is a summarised list:
Process and disclosure failures. (Governance)
Boundary and isolation failure (Technical)
Control Flow and Safety failure (Model risks)
Attack surface & misuse (Threat actors and Competitive narratives)
User harm and trust impact.
Here is the issue. We spend a lot of time on these. Time that could be spent helping people understand AI that vendors no not want to educate the public in.
I shall now resume screaming into the void where apparently all tickets that pre-empt major impacts live.
APS AI Plan
3:34 PM
Welcome to the segment of “What fresh hell is this?”. I will have an open mind reviewing this but my expectations are ever so low.
EDIT: 9:37PM. I no longer have faith in the government and its various spin offs and committees. At the detriment to the people’s safety it frequently mentions, a majority of the review and our findings prove one major and frightening artifact.
These people have no idea how to use AI. The least you could do is set up AU spelling.
The other obvious and expected one: This is not supportive of the people but a cabal and follows our predictions accurately. May consider making our predictions a bingo card for 2026.
