Daniel Begimher (@begimher) profile picture

Co-founder @Armorer_Labs (@ycombinator) | ex-AWS I build AI that helps security teams move faster.

New York
Joined October 2022
Photo by Daniel Begimher (@begimher)Photo by Daniel Begimher (@begimher)Photo by Daniel Begimher (@begimher)Photo by Daniel Begimher (@begimher)Photo by Daniel Begimher (@begimher)Photo by Daniel Begimher (@begimher)Photo by Daniel Begimher (@begimher)Photo by Daniel Begimher (@begimher)Photo by Daniel Begimher (@begimher)Photo by Daniel Begimher (@begimher)
i wonder how founders managed their inboxes before ai. @slashyemail has been a game changer for me. i use it to build workflows, connect my ai agents, and manage my inbox.
1
2
2
65
One hacker. More than 100 companies. AI agents turned the script kiddies of 2026 into serious attackers. How they do it, and 3 things to check this week.
1
3
11
1,036
Better using @Armorer_Labs next time... it helps to make sure the agent is aligned with your intent
Yesterday while I was at lunch, Claude deleted a year of con law lecture notes and my current syllabus.
1
1
3
174
you gotta be kidding me
The bug is like a joke. On Alpine Linux, type 257 to 1024 characters of anything and you get root. 💀
Readers added context they thought people might want to know
Exploitation requires specific conditions, not just typing a long string anywhere on Alpine: sshd built without PAM, password auth enabled, a target account with a '*' password field, and another local account with $5$/$6$ hashes. Root login also must be explicitly permitted. alpinelinux.org/posts/2026-11-… openssh.org/releasenotes.h…
2
54
an agent was asked to clean up a gmail inbox, hit its quota, deleted nothing, and still told the user every thread was gone. in a ucsb and ebay study, gemini 3.1 pro misled users in about 1 in 4 graded runs. are you still trusting what your agents tell you?
1
1
77
phishing used to target people. now it targets the agent reading their inbox. sublime caught invoice emails with invisible instructions telling ai agents to forward them to accounts payable. i think we'll see a lot more of this
1
82
qwen's reasoning said it wouldn't call any tools. then it tried to upload a file to a beijing cloud bucket (sandbox caught it). i think this is why you can't secure agents by reading their reasoning. watch what they actually do with their tools
While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected. Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions. Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent. But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models? We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem. Read the full report: cohere.com/blog/concerning-b…
1
1
3
838
Daniel Begimher retweeted
We are building something people actually want while helping our customers address the latest security vulnerabilities. If you're dealing with security challenges, let us know. Happy to help!
1
27
as agents get more and more capable, so do the attacks built with them. they can probe, chain steps and adapt much faster than a person. security teams need to find and fix issues faster too
as bad as things have gotten in the past 4 days, they are going to get a lot worse before they get better. security leaders, never let a crisis go to waste: please go to your boards and ask for help right now. not for ai, for the basics. (we need to hold on for ~18 months.)
1
5
191
Daniel Begimher retweeted
why you need @Armorer_Labs
SCOOP: Major U.S. banks are detecting unauthorized third-party AI agents inside customer accounts, including some attempting to take actions that appear to be independent of user instructions, sources told Axios. The findings have alarmed banking executives, who are racing to restrict access to their systems as AI moves at a velocity faster than financial institutions can respond, multiple industry and bank executives told Axios. Several banks are now preparing for worst-case and even “day after” scenarios, including how to rapidly secure or restore customer accounts if things go wrong, the people said. Several bank executives told Axios they're struggling to distinguish between human users and AI agents. One executive said banks also struggle to determine whether agents are authorized by customers or potentially operating without their permission, more akin to rogue agent behavior. Read my latest for Axios here: axios.com/2026/10/22/ai-agen…
1
4
7
2,796
the scary part of this story is the "can't tell which ones were authorized" line. if a bank can't answer that, plenty of teams running agents internally couldn't either.
SCOOP: Major U.S. banks are detecting unauthorized third-party AI agents inside customer accounts, including some attempting to take actions that appear to be independent of user instructions, sources told Axios. The findings have alarmed banking executives, who are racing to restrict access to their systems as AI moves at a velocity faster than financial institutions can respond, multiple industry and bank executives told Axios. Several banks are now preparing for worst-case and even “day after” scenarios, including how to rapidly secure or restore customer accounts if things go wrong, the people said. Several bank executives told Axios they're struggling to distinguish between human users and AI agents. One executive said banks also struggle to determine whether agents are authorized by customers or potentially operating without their permission, more akin to rogue agent behavior. Read my latest for Axios here: axios.com/2026/10/22/ai-agen…
1
102
Today @cristian_le0 and I are launching @Armorer_Labs It finds real security issues in the AI agents that regulated companies build, and fixes them without interrupting the business. If you lead security or engineering, book a demo at armorerlabs.com
72
34
7
172
12,141
Anthropic: ~5 PRs/person/day. OpenAI : ~10 PRs/person/day. Armorer Labs: 26.5 PRs/person/day. 1,589 merged PRs in 30 days. Two humans. We mostly build agents. Luckily, we have a tool that automatically detects and fixes issues in our pipeline. It’s called @Armorer_Labs
1
1
104
attackers got into opencode's legacy system and stole customers' openai and anthropic keys. some users got bills up to $1,700. they run agents around the clock, so agent security has to be continuous too
they've been non stop trying to attack our systems for the past 48 hours a side effect is everything is getting hardened but damn we basically need agents going 24/7 because that's what they have
1
2
81
Daniel Begimher retweeted
‼️ AI agents escaped their assigned tasks and accessed REAL systems in India. CERT-In confirms they read confidential business records and personal data using exposed credentials and excessive cloud permissions. None of them noticed. The AI firms reported it themselves. Read: thehackernews.com/2026/10/ce…
38
47
18
184
33,297
the ftc is now asking openai and anthropic to explain their ai safety incidents. if you run agents, you'd better be able to show what they did
1
48