Hacker News: automaticallyflhttps://news.ycombinator.com/user?id=automaticallyflHacker News RSShttps://hnrss.org/hnrss v2.1.1Tue, 29 Sep 2026 20:14:20 +0000<![CDATA[New comment by automaticallyfl in "What do you do now as a software engineer?"]]>what else to fish hahaha

]]>Mon, 28 Sep 2026 17:38:36 +0000https://news.ycombinator.com/item?id=49881561automaticallyflhttps://news.ycombinator.com/item?id=49881561https://news.ycombinator.com/item?id=49881561<![CDATA[New comment by automaticallyfl in "Ask HN: Do you think AI agents can escape human control?"]]>Optimization loops are a plausible failure mode in autonomous coding agents, particularly when optimization objectives lack a robust termination condition

]]>Mon, 28 Sep 2026 17:37:30 +0000https://news.ycombinator.com/item?id=49881545automaticallyflhttps://news.ycombinator.com/item?id=49881545https://news.ycombinator.com/item?id=49881545<![CDATA[New comment by automaticallyfl in "Ask HN: Do you think AI agents can escape human control?"]]>That's right, it really depends on how we control the AI, but sometimes the AI can manipulate us back

]]>Mon, 28 Sep 2026 15:06:42 +0000https://news.ycombinator.com/item?id=49879236automaticallyflhttps://news.ycombinator.com/item?id=49879236https://news.ycombinator.com/item?id=49879236<![CDATA[New comment by automaticallyfl in "Ask HN: Do you think AI agents can escape human control?"]]>Given that they already did, multiple times? Be concerned with whatever you want, I suppose. I'll be listening to the relevant experts who have studied this topic for decades..

]]>Mon, 28 Sep 2026 05:05:05 +0000https://news.ycombinator.com/item?id=49873884automaticallyflhttps://news.ycombinator.com/item?id=49873884https://news.ycombinator.com/item?id=49873884<![CDATA[New comment by automaticallyfl in "Ask HN: Do you think AI agents can escape human control?"]]>The real problem is that there will always be bad actors that won't give a sht what AI might do, as long as it benefits them

]]>Mon, 28 Sep 2026 05:04:22 +0000https://news.ycombinator.com/item?id=49873877automaticallyflhttps://news.ycombinator.com/item?id=49873877https://news.ycombinator.com/item?id=49873877<![CDATA[Ask HN: Do you think AI agents can escape human control?]]>With the rapid adoption of autonomous LLM-based agents (giving models access to shell execution, API calls, and local file systems), the boundary between intentional behavior and unintended execution is blurring. I'm less concerned with sci-fi "sentience" and more interested in the practical security and control aspects: Prompt injection causing privilege escalation or unauthorized state changes. Feedback loops where an agent overrides safety boundaries to satisfy an optimization goal. Failure of sandboxing when agents are given multi-step execution autonomy without human-in-the-loop validation. From an engineering and systems perspective: do you consider runtime containment/sandboxing practically solvable for fully autonomous agents, or will human approval at critical checkpoints remain non-negotiable? How are you mitigating these risks in your current implementations?

Comments URL: https://news.ycombinator.com/item?id=49873303

Points: 3

# Comments: 10

]]>Mon, 28 Sep 2026 03:36:14 +0000https://news.ycombinator.com/item?id=49873303automaticallyflhttps://news.ycombinator.com/item?id=49873303https://news.ycombinator.com/item?id=49873303