Home About Black Hat Contact Essays Fireside Chats My Take News Alerts Q&A Reels RSAC Videocasts ☰
 

Q & A

 

SHARED INTEL Q&A: PKI’s unfinished business—’digital passports’ for content, models and agents

By Byron V. Acohido

As if keeping track of machine identities wasn’t hard enough.

AI agents are now arriving by the thousands — and most enterprises are just handing them borrowed credentials and hoping for the best.

Meanwhile, the cryptographic infrastructure asked to absorb these threats faces a hard regulatory countdown requiring digital certificates — the credentials securing every machine connection — to rotate eight times more often than they do today.

Related: Certificate expiration is speeding up

That same foundation now faces three converging demands at once: vouch for AI agents, authenticate synthetic media and survive the coming migration to quantum-safe cryptography — all on overlapping deadlines

SHARED INTEL Q&A: AI retrieval systems can still hallucinate; deterministic logic offers a fix

By Byron V. Acohido

AI hallucination is still the deal-breaker.

Related: Correcting LLM hallucinations 

As companies rush AI into production, executives face a basic constraint: you cannot automate a workflow if you cannot trust the output. A model that fabricates facts becomes a risk exposure. CISOs now have to explain this clearly to boards, who expect assurances that fabrication risk will be controlled, not hoped away.

Earlier this year, retrieval-augmented generation, or RAG, gained attention as a practical check on hallucination. The idea was straightforward: before answering, the model retrieves grounding material from a trusted source and uses that to shape its response. This improved reliability in many early use cases.

But first-generation RAG had a hidden weakness. A major academic study (“RAGTruth”) showed that even when RAG retrieves accurate source material, AI systems can still misstate it or draw the wrong conclusion. The research comes from the ACL Anthology, the leading global library for peer-reviewed AI language research.

AUTHOR Q&A: New techno-thriller ‘The Virus’ simulates an AI malware outbreak gone global

By Byron V. Acohido

Eddy Willems has been a steady, pragmatic voice in cybersecurity for decades — known for breaking down complex threats in ways real people can understand.

Related: AI fueling disinformaton

With The Virus, he tries something new: a fast-paced techno-thriller that fuses autobiography with speculative fiction. The result reads part memoir, part cautionary tale, part simulation of how a global cyber outbreak might unfold.

The story toggles between threat briefings and the fictional rise of “Doomware,” a novel virus that spirals from localized disruption into full-blown global crisis. Willems casts himself as a lightly fictionalized version of his real-life role, offering insight through scenes that draw on years of direct frontline experience.

In today’s threat landscape — shaped by AI acceleration, ransomware-as-a-service, and politicized disinformation — The Virus feels eerily plausible. Its fictional twists are anchored in real-world behaviors and known vulnerabilities. Because Willems never strays too far from facts, the book functions as both gripping story and subtle cybersecurity literacy. It doesn’t preach, but it lands with urgency.

NEW TECH Q&A: Why Data Bill of Materials (DBOM) is surfacing as a crucial tool to secure AI

By Byron V. Acohido

Enterprises hustling to embed AI across their operations came to an uncomfortable realization in 2025: they lost track of the data powering those systems.

Related: The case for SBOM

Few paused to map where sensitive data lived or how it moved. That oversight is now surfacing in audits, breach incidents, and regulatory pressure.

According to field research from Bedrock Security, most IT and security leaders still lack foundational visibility into the datasets fueling training and inference.

“You can’t govern retroactively,” says Bruno Kurtic, co-founder and CEO of Bedrock Data. “You need controls before the model runs, not after.”

Last Watchdog caught up with Kurtic to understand what this shift means in practical terms. Here’s an edited version of that conversation.

SHARED INTEL Q&A: This is how ‘edge AI’ is forcing a rethink of trust, security and resilience

By Byron V. Acohido

A seismic shift in digital systems is underway — and most people are missing it.

Related: Edge AI at the chip layer

While generative AI demos and LLM hype steal the spotlight, enterprise infrastructure is being quietly re-architected, not from the cloud down, but from the silicon up.

As AI use cases move from experimentation to deployment, a new layer of compute is taking shape. Behind the scenes, semiconductor companies like Infineon are already adapting embedded systems and edge architectures to handle persistent inference workloads — AI that must run in real time, in the field, at the device level.

NEW TECH Q&A: Start-up Identient debuts reimagined AI copilots trained on experts’ insights

By Byron V. Acohido

Cybersecurity has always been a moving target. But AI has shifted the center of gravity.

Related: The workflow cadences of Gen AI

Over just the past two years, we’ve watched decision cycles compress, incentive structures change beneath us, and the mechanisms for validating expertise start to break down in real time.

Threat actors are already using AI to scale speed and improvisation. Defensive teams are retooling under pressure. Meanwhile, the advisory layer — analysts, consultants, research channels — has struggled to keep up with risks evolving faster than their traditional frameworks allow.

At Last Watchdog, we’ve noted how even well-established decision models begin to falter once AI-driven ambiguity and tempo are introduced. That’s why I took interest in a new approach from Seattle-based Identient. Led by founder and CEO, Steve Tout, the company is launching what it calls the first marketplace for “verified digital twins”: AI counterparts trained only on authenticated expert material and constrained by provenance and governance.

SHARED INTEL Q&A: API gaps expose AI fault lines — an urgent call for hygiene, active monitoring

By Byron V. Acohido

The race to deploy GenAI far and wide has intensified enterprises’ reliance on APIs — most of which remain poorly understood and underprotected.

Related: Mistaking AI pattern matching for wisdom

This reality is highlighted in a just-released Salt Security survey of 250 security and IT leaders which found nearly half (48%) reporting API security concerns slowing down their organizations’ AI adoption.

It tracks. Generative AI relies on connecting models to internal data and functions — often through complex webs of APIs. But most companies still don’t have full visibility into their APIs, let alone control over how they behave in real time. Shadow APIs, zombie APIs, weak governance — these gaps are translating into AI blockers.

The recent OneLogin breach is a case in point. Attackers used a compromised API key to access internal apps and customer data — and the intrusion went undetected for weeks. Traditional tools like WAFs, gateways, and static scans missed it. They weren’t built to track today’s fast-changing API traffic or catch business logic abuse in production.