How much do you trust AI agents?
With the advent of clawdbots, it's as if we've all lost our inhibitions and "put our lives completely in their hands."
I'm all for delegating work, but not giving them too much personal/sensitive stuff to handle.
I certainly wouldn't trust something to the extent of providing:
access to personal finances and operations (maybe just setting aside an amount I'm willing to lose)
sensitive health and biometric information (can be easily misused)
confidential communication with key people (secret is secret)
Are there any tasks you wouldn't give AI agents or data you wouldn't allow them to access? What would that be?
Re. finances – Yesterday I read this news: Sapiom raises $15M to help AI agents buy their own tech tools – so this may be a new era when funds will go rather to Agents than to founders.
Replies
the real problem nobody here is talking about is approval fatigue. you set up all these guardrails and least privilege rules, agent asks for permission 30 times a day, and within a week youre just hitting approve on everything without reading it. saw someone above mention this exact thing.
ive been running a coding agent basically 24/7 for months and the stuff that actually keeps it safe isnt permissions or sandboxes. its making destructive actions harder than constructive ones. like using trash instead of rm, writing drafts instead of sending emails directly, committing to branches instead of main. the friction is the guardrail, not some approval popup you stop reading after day 2.
the people saying "never give agents access to X" are thinking about it wrong imo. its not about what you give access to, its about whether the bad outcome is reversible. read access to your bank account? who cares. write access to send a wire? yeah thats different. but most people draw the line at "personal = off limits" which doesnt actually map to risk at all.
minimalist phone: creating folders
@umairnadeem but with this... not many things are reversible. When something once happens, it happens and cannot be returned; only you can "lighten" the consequences (and you have to have in mind possible damages and risks).
Depends on the task. I trust completely when it comes to classifying my inbox.
minimalist phone: creating folders
@salah_oukrim My inbox is something I wouldn't like to give access to anybody and anything :D too sensitive :D
the real problem nobody here is talking about is approval fatigue. you set up all these guardrails and confirmation prompts and for the first week you actually read them. then by week two youre just hitting approve on everything because the agent asks permission 40 times a day and you stop caring.
i run coding agents pretty much nonstop and ive caught myself approving stuff i didnt even read twice. not because i trust the agent but because the friction of constantly reviewing breaks your flow. so you end up in this weird middle ground where you technically have human oversight but practically dont.
honestly the answer isnt more guardrails or less trust. its better scoping. give the agent a tiny sandbox where it literally cant do damage even if it goes rogue, and let it go wild in there. trying to micromanage every action just means you eventually stop managing any of them.
Murror
Building Murror has made me think about this from both sides. I am building an AI that people trust with something really personal, their loneliness and emotional state. So the question of trust is not abstract for me.
My take is that trust with AI agents is earned the same way it is with people: through consistency and transparency about what they are actually doing. The fear is not the autonomy itself, it is not knowing what happened.
For Murror I think about it this way: the AI should be a safe space, not a black box. Users need to feel like they understand what it does with what they share. That is the design challenge I care about most right now.
minimalist phone: creating folders
@astrovinh for any agent, I would like to have a middle step: approval process – dunno, I am a kind of paranoid :D
AI is great for automation and productivity, but it still needs boundaries.
I would never give AI full access to my finances, sensitive health or biometric data, or confidential conversations. Those things require human trust and control.
AI should assist decisions, not own them.
minimalist phone: creating folders
@abhay_donde You speak with my "trust" language ;)
@busmark_w_nika Haha maybe a little. But I think trust is the real foundation of any technology. Tools can do amazing things, but some things still belong in human hands. AI should help us think better, not replace our judgment...
minimalist phone: creating folders
@abhay_donde Let's see at what extent we can collaborate with AI :)
@busmark_w_nika Exactly… the real skill is knowing where to draw the line 😉
I think AI is great for automation and assistance, but there are clear boundaries. I wouldn’t give it full access to personal finances, biometric data, or private conversations. Those are areas where a mistake, breach, or misuse could have serious consequences.
For me, AI works best as a tool that suggests and helps, not something that has complete control over sensitive parts of my life.
minimalist phone: creating folders
@emmanuel_afolabi Definitely, we shouldn't become slaves of AI. (the thing that happened with us and social media)
Murror
Coming at this from a slightly different angle. I build Murror, an AI app for emotional support and loneliness, so trust isn't just a nice-to-have for us, it's basically the whole product. If people don't feel safe being vulnerable with it, there's nothing there.
What I've noticed is that trust with AI in emotional contexts is earned really slowly and lost really fast. One weird or cold response and the person closes off. It's different from a productivity tool where a mistake is just annoying.
I think the harder question isn't "how much do you trust AI" but "does the AI know what it's holding." A lot of agents don't seem built with any awareness of how sensitive the context actually is. That gap worries me more than the capability questions.
minimalist phone: creating folders
@astrovinh But even with that, we can trust only partially, it is the same as humans, we cannot trust 100% :)
Murror
the financial and task stuff i'm fairly comfortable delegating. but the thing i'm most cautious about is emotional context. i build in the mental health space and the one thing i keep coming back to is that AI can be very convincing while still being completely off about what someone actually needs.
for sensitive health or personal stuff, i think the risk isn't just misuse of data. it's the AI confidently reflecting something back to a vulnerable person that's just wrong. that's a harder problem than access control.
so my line is: i trust agents with information. i don't trust them with interpretation of people's emotional states, at least not without a lot of care in how it's built.
minimalist phone: creating folders
@astrovinh But I think AI can spot some behavioural patterns that can reveal the problem. I wouldn't neglect it :)
MacQuit
As a developer, I use AI agents daily for coding — and honestly they've changed how I work. But my trust level really depends on one thing: is the action reversible?
Writing code? Sure, let the agent go wild. I can always review the diff and revert. Deploying to production? Hard no without me looking at it first. Sending an email to a client? Absolutely not.
My rule of thumb after 10+ years of shipping software: the more irreversible an action is, the more human oversight it needs. AI agents are incredible at drafting, exploring, and iterating. But the "commit" moment — whether it's pushing code, sending money, or publishing something — that should stay with a human.
What worries me most isn't the AI making mistakes (it will, that's fine). It's the speed at which mistakes can cascade when there's no human checkpoint. One bad API call from an agent with too many permissions, and you've got a real mess on your hands.
I think the sweet spot right now is: AI does 90% of the heavy lifting, human approves the final 10%. Not because the AI can't do it, but because we haven't built the trust infrastructure yet. We'll get there, but rushing it would be unwise.
minimalist phone: creating folders
@lzhgus In that case, you need to think about possible worse scenarios in advance, because when the thing is done, it cannot be undone, you can only limit damages by prevention or earlier intervention.
I guess it would depend on what it is access. I have a SaaS that accesses (read only) my bank data, but I want an automated process to add my transactions for the SaaS as it is a tax preparation software. I don't have an issue with that. The other day I used AI the other day to research some medical symptoms I was having. When I spoke with the doctor it was what AI already had told me, but AI asked the same questions the doctors had asked. I don't have a problem with that as I am trying to figure out what is going on and what I can do about it.
when it comes to dividing topics that I am doing research on, I make sure that I pose questions and hypothesis' I remain neutral and you have to very deliberate in this approach so the AI doesn't build responses on any bias's you may have. this approach will deal more in facts than a leaning direction
minimalist phone: creating folders
@david_sherer Even access to reading bank data is too sensitive for me. 😅
@busmark_w_nika was for my brother too. :)