Oh my god, OpenAI reinvented Recall, but for macOS.

[mttaggart@infosec.exchange]

Lmaooo

Using Chronicle increases risk to prompt injection attacks from screen content. For instance, if you browse a site with malicious agent instructions, Codex may follow those instructions.

One wonders how many things it will OCR into a prompt injection. Utter clownshow of a technology.

[jtig@infosec.exchange]

@mttaggart "Just make sure you don't read anything malicious"

[kate@hai.z0ne.social]

@mttaggart@infosec.exchange AI is the skibidi 6 - 7 of the 2020s.

[smilingdemon@mastodon.art]

@jtig @mttaggart "only read things you have read before and know to be safe."

[peterbrown@mastodon.scot]

@mttaggart “opt-in”

[kluthulhu@infosec.exchange]

@mttaggart Counting the days until platforms like hackthebox tell you to "deactivate Recall/Chronicle before booting the VM"

[make@mastodontti.fi]

@mttaggart oh dear, poor Bobby Tables here we go again...

[wikiyu@infosec.exchange]

@grumpydad @mttaggart microsoft with recall, and ... why do you assume they thought its good?
Evil people do evil things.

they want all our personal data.
not to make good

[q_observations@mstdn.social]

@mttaggart *tech-no-logic

[dazo@infosec.exchange]

@mttaggart I just wonder .... How readable must those instructions be? Does it need to be visible at all (think, white-on-white, div tag with display:none type of scenarios) ... or does it need to have some contrast? If so, how much? Can it be blended into a background image?

I'm preparing some now

[buherator@infosec.place]

@mttaggart