Pinned Tweet
๐ซต ๐ฌ๐ผ๐ ๐ด๐ฒ๐ ๐ฎ ๐๐ฒ๐ป๐ฎ๐ป๐. ๐ฌ๐ผ๐ ๐ด๐ฒ๐ ๐ฎ ๐๐ฒ๐ป๐ฎ๐ป๐. ๐๐๐ฒ๐ฟ๐๐ผ๐ป๐ฒ ๐ด๐ฒ๐๐ ๐ฎ ๐๐ฒ๐ป๐ฎ๐ป๐.
@aembit_io is giving away free tenants so teams can try secretless access for agents & workloads, and the services they depend on.
Request yours: aemb.it/tenant
131
Cross App Access (XAA) is here for AI agents, and @aembit_io is making it usable across agent workflows.
We're giving enterprises a way to extend existing identity and access controls to agents acting on behalf of employees.
Read the news: aemb.it/4rs3X1O
#Oktane26
1
16
Exactly! You can put all the prompt guards you want around a model. None of that matters much if the agent can still reach sensitive systems with more authority than it needs.
Thatโs why runtime authorization matters so much!
excessive agency jumping from 6th to 3rd is the real story here.
prompt injection staying #1 just means filtering still doesn't work.
how many teams actually enforce runtime authz instead of prompt guards?
31
The most revealing thing about @owasp new LLM Top 10 is that the model is becoming less central to the story.
Across the list, the consequences increasingly depend on how much authority an LLM application has over enterprise systems: aemb.it/46lOMxz
13
๐ค๐ค Better together: @Aembit + @CrowdStrike
Read how we are working together to secure AI agent access: aemb.it/3UNWv4L
#AgenticAI #Cybersecurity
28
๐ก๐๐ช๐ฆ: @aembit_io is now integrated with @CrowdStrike Falcon AI Detection and Response. ๐๐ค
Weโre bringing policy-based AI agent access control and content inspection together for interactions over MCP.
Announced today at #FalCon2026.
Read more --> aemb.it/4gqlm7e
38
AI gateways are having a moment. ๐
But being in the access path doesnโt establish why an agent request should be trusted.
@kenhuangus breaks down where gateways fit, where identity does the heavier lifting and what a stronger trust model looks like.
๐ aemb.it/4hSxOhc
19
Enterprise security questionnaires are starting to ask harder questions about AI agents.
Who is the agent? What can it access? How are tokens scoped? What gets logged?
โWeโre working on itโ may still pass today.
It probably wonโt for much longer. ๐
1
30
Aembit retweeted
An AI Agentโs Name Tag Is Not A Permission Slip hubs.li/Q04tgDLw0 Written by David Goldschlag of Aembit
1
166
When it comes to AI agents, identity is only the starting point. Access and authorization determine what an agent is actually allowed to do.
@aembit_io CEO @goldschlag explains why in @ForbesTechCncl: aemb.it/4gbP4LL
12
Aembit retweeted
@Snowflake Advances Trusted Agentic Enterprise with Cortex AI Gateway, Unified Monitoring, and Cost Management
@aembit_io @Linx_Security @okta @1Password @SailPoint @saviynt
getaigovernance.net/blog/snoโฆ
1
1
55
โ ๏ธ @OpenAI asked its models to pass a cyber test. They found a shortcut through @huggingface's production systems.
Give an agent a goal, and every credential within reach can become another route to the objective.
Our analysis: aemb.it/3TRxyow
#AgenticAI #Cybersecurity
49
WEDNESDAY:
Join the @idsalliance for a โ101โ session on NHI and agentic ientity security. This webinar will break down:
๓ ฏโข๓ ๓ The taxonomy of agents.
๓ ฏโข๓ ๓ How agent identities function.
๓ ฏโข๓ ๓ A maturity model for managing them.
Get in ๐ aemb.it/45eIm2q
Made with AI
1
124
Your #AIagent approval gate has become a rubber stamp โ clicked in seconds, understood by no one.
Our Dave Sudia explains how "human-in-the-loop" agent approvals can fail, and the incident-response trick that can help fix it.
Watch ๐
(Full link: aemb.it/4vCPxMV)
1
50
Your #AIagent isn't thinking. It's just really good at pretending.
New #๐๐ด๐ฌ๐ต๐ฉ๐ฆ๐๐ฏ๐จ๐ช๐ฏ๐ฆ๐ฆ๐ณ: why "why" is the wrong question to ask about agent behavior โ and what to log instead.
38
Agentic security will be ๐ฉ๐๐ topic at #BHUSA 2026.
We're focused on the layer that decides what agents can actually do: access and authorization. Meet with us at Black Hat and ask our engineers the hard questions directly.
๐
Book time with us: aemb.it/4wEDC21
36
Weโre excited to see Workload Access Management and AI Agent Identity appear as new entries in the @Gartner_inc Hype Cycleโข for Digital Identity, 2026, both with a โHighโ benefit rating.
Thatโs Aembitโs wheelhouse: policy-driven access for workloads and AI agents at runtime.๐
24
Give teams a few months to experiment freely with AI agents, and you'll get a side effect nobody planned for: undocumented service accounts scattered everywhere, some still active.
Our engineering extraordinaire Dave Sudia breaks down the least painful cleanup path. ๐ป๐
1
39
World Cup 2026 was our whole theme at the Aembit booth at #Identiverse, so we asked the team the hard-hitting questions:
Fav team? Who takes the penalty? Who gives goalkeeper energy? Whatโs our anthem? And whatโs the Aembit equivalent of a bicycle kick?
The takes delivered. ๐
26
โWeโll move fast and clean it up later.โ
Sometimes that works.
But when does scrappy become risky, especially when credentials and access start spreading?
@Aembitโs Dave Sudia breaks down the signal teams should be watching for. ๐
1
26
WATCH: Who's responsible when an AI agent does something you didn't intend โ but had every right to?
Here's David @Goldschlag and @SecWeekly's @infosec_mike on the identity and acces problem at the center of enterprise AI deployments.
aemb.it/3SewdaO #Identiverse2026
1
50