DEV Community: Security Expert

I think the AI vs AI cybersecurity debate is missing the real risk

Security Expert — Thu, 02 Apr 2026 12:15:00 +0000

Everyone’s talking about attackers using AI vs defenders using AI.

But after working closely with teams at Periscope Technologies Inc, I’m starting to feel like that’s not where things are breaking.

What we’re seeing more often:

AI inside companies already has:

• Access to sensitive data
• Ability to trigger workflows
• Decision-making power

But very little verification around:

👉 What it’s doing
👉 Why it’s doing it
👉 Whether it should be doing it

Traditional security made sense when:

• Users were human
• Behavior was predictable
• Access was controlled

AI doesn’t fit that model at all.

Feels like we’re focusing heavily on external AI threats…

While ignoring a new category of risk:

Unverified AI execution inside systems

Curious if others here are seeing this?

Or is the focus still mostly on external threats?

periscope-tech.com

$14.5B wiped out in a day after an AI shock overreaction or real signal?

Security Expert — Mon, 30 Mar 2026 10:47:32 +0000

A $14.5B drop in cybersecurity stocks triggered by an AI-related event is being called market volatility.

Not sure I agree.
Feels more like a signal.

AI is moving faster than security can keep up.

And most companies are already using AI internally for:

• Data access
• Decision-making
• Workflow automation

But very few have real control or visibility into:

👉 What it’s doing
👉 What data it’s touching
👉 What decisions it’s making

Working with teams at Periscope, this gap shows up a lot.

Not in adoption.

In control.
Curious how others see this:

Is this just hype…
or an early warning?

https://periscope-tech.com/periscope-secure-access-platform

What basic security checks do you actually follow every week?

Security Expert — Fri, 27 Mar 2026 10:58:16 +0000

In many setups I’ve seen, the real issues aren’t complex — it’s small things getting missed:

Old user/vendor access still active
Backups assumed to be working
Critical updates delayed
Alerts no one checks
No clear owner if something breaks

Individually, these don’t feel urgent.
But that’s usually where problems start.

Do you follow a weekly checklist for this?
Or is it more reactive in your environment?

https://periscope-tech.com/periscope-secure-access-platform

Deploying Agentic AI Safely: Lessons from This Week in AI

Security Expert — Mon, 23 Mar 2026 11:24:45 +0000

This week, agentic AI made headlines:

Kill-switches being implemented
Agents treated like employees with identities
Legal challenges emerging
Dedicated infrastructure for execution

Most deployments fail because control and monitoring aren’t built-in.

At Periscope, we deploy secure, production-ready agentic workflows. Try 1 workflow risk-free for 2 weeks, with monitoring, guardrails, and permissions baked in.

https://periscope-tech.com/agentic-factory

Why Replacing Hospital Systems Fails (And What Actually Works)

Security Expert — Fri, 20 Mar 2026 11:00:55 +0000

Most hospitals already have the systems they need.

EHR. CRM. RCM. Contact center.

The real problem?
They don’t work well together.

That’s where most “AI transformations” go wrong —
they try to replace instead of optimize.

A more practical approach:

Embed AI into existing workflows across
intake → scheduling → care → billing.

Instead of adding complexity, it removes friction.

What changes:

Intake becomes faster

Admin workload drops

Revenue leakage is reduced

No rip-and-replace.
No operational disruption.

Start with one workflow. Measure impact. Expand from there.

Full approach:
https://periscope-tech.com/agentic-factory

SaaS is quietly being replaced by Agent-as-a-Service (AaaS) — are we ready for this?

Security Expert — Tue, 17 Mar 2026 16:26:33 +0000

This week had some interesting signals:

NVIDIA pushing infrastructure for AI agents

Alibaba launching enterprise agent platforms

Feels like we’re moving from SaaS → systems that actually execute tasks.

Not just dashboards, but AI that:

updates systems

triggers workflows

makes decisions

The interesting part:

The biggest challenge isn’t capability.
It’s control.

If an AI agent takes the wrong action inside your stack, the risk is very real.

Curious how people here are thinking about this:

Are you experimenting with agents yet?
Or does this still feel too early / risky?

periscope-tech.com

Fake Claude Code Download Pages Are Delivering Infostealer Malware

Security Expert — Mon, 16 Mar 2026 11:36:59 +0000

As AI development tools continue to grow in popularity, attackers are beginning to exploit their ecosystems.

A recent campaign is distributing fake Claude Code download pages that mimic legitimate installers.

The attack relies on sponsored search advertisements, meaning developers searching for Claude Code may encounter the malicious page before the real one.

If the installation command is copied and executed, the malware deploys infostealers capable of capturing:

• Browser credentials
• Session cookies
• API tokens
• Crypto wallet data

This is another sign that AI tooling ecosystems are becoming a new supply-chain attack vector for developers.

Full threat analysis:
https://blog.periscope-tech.com/new-threat-alert-fake-claude-code-downloads-delivering-infostealers/

cybersecurity #devops #infosec #ai #security