Mike Mosher

Cybersecurity Predictions for 2026: The Future of Digital Threats Cybersecurity experts discuss 2026 predictions, highlighting the rise of AI-driven threats and the need for advanced security measures to combat risks.

Dark Reading reports that for 2026, security must shift from prevention to resilience. AI attacks are too fast for old methods. If you aren't spending 30% of your budget on recovery, you are behind. Focus on MTTR to protect shareholders.

www.darkreading.com/threat-intel...

Me: “Polly, you can’t say that! Bad parrot! Say you’re sorry.”

🦜 Polly: *Awwkk* “Bad parrot” *Awwkk* “Sorry” *whistle* 🦜

Me: “Okay, Polly, don’t do it again.”

🦜 Polly: *Awwkk* “Do it again…do it again!” *Awwkk* *whistle* 🦜

#ai #agents #severance #engineering #security #sdd | Michael Mosher Thinking about "Severance" and AI Agents. If you've seen Severance, you know the horror of the "Innie." Waking up at work with no memory of the past. It’s a dystopian nightmare for humans. But it’s ...

www.linkedin.com/posts/michae...

The Amnesiac Intern Why AI Agents Need External Memory

open.substack.com/pub/northwoo...

We need to stop waiting for infinite context windows and start building better harnesses.
Give your digital interns a notebook that survives the reboot. 4/

#AI #AgenticWorkflows #Severance #Security #SDD

This is huge for Security.
Imagine a Red Team agent finding a bug. If it doesn't leave a perfect map (External Memory) for the next agent, you waste tokens and miss fixes.
The "Spec" isn't just instructions—it's the Agent's long-term memory. 3/

I read Anthropic's new paper on "Long-Running Agents" and the solution is clear:
It's not just about smarter models (better interns).
It's about better Context (external memory).
You need a "Manager" agent to leave a detailed "To-Do" list for the "Worker" agent. 2/

If you've seen Severance, you know the horror of the "Innie"—waking up at work with no memory of the past.
That's exactly how we treat AI Agents.
We spin them up, they work, we kill them.
The next agent starts with zero memory of what the previous one did.
It's the Amnesiac Intern problem. 🧵 1/

Now security departments are going to have to figure out how to control AI agents in the hands of every knowledge worker?

Tell me what you think...
(3/3)

I mean, do we want our knowledge workers using AI to create and deploy AI agents? Companies expend a lot of effort to keep these knowledge workers from clicking on phishing emails or sharing sensitive information when they set up their out of office notifications. (2/3)

Google Workspace Updates: Now available: Create AI agents to automate work with Google Workspace Studio

Google has announced Google Workspace Studio to allow enterprise Google users to create and share AI agents.

Google is continuing to pull ahead in AI, but is this really a good thing? (1/3)

workspaceupdates.googleblog.com/2025/12/work...

Watching a really interesting presentation by Dr Chase Cunningham at the Cloud Security Alliance Zero Trust Summit. He’s making a strong case for Cybersecurity Tax Credits as a means to incentivize better practices, increased investment, and risk reduction. #cybersecurity

Big news in the Identity & Access Management (IAM) market:

ServiceNow (NYSE: NOW) is acquiring Veza, giving them a solid entry into the Identity Governance & Administration space, and validating the importance of this capability in today's security and tech stack.

#cybersecurity #servicenow #IAM

The AI Paradox: Why Hyper-Efficiency Could Mean More Human Work How a 19th-century economic theory can inform our thinking about AI and the future of work.

My latest on Substack:
open.substack.com/pub/northwoo...

#AI #FutureOfWork #JevonsParadox #ArtificialIntelligence #Leadership #TechTrends #Innovation #Economics #CloudComputing

Hackers Leverage Evilginx to Undermine MFA Security Mimicking Legitimate SSO Sites

Hackers Leverage Evilginx to Undermine MFA Security Mimicking Legitimate SSO Sites

ChatGPT is down worldwide, conversations dissapeared for users OpenAI's AI-powered ChatGPT is down worldwide, and the reason is unclear.

ChatGPT is down worldwide, conversations dissapeared for users

Chinese Maker Behind Most of World’s Bitcoin Miners Has Been Focus of US National Security Probe Investigators have sought to assess whether Bitmain Technologies’ products pose risks of espionage or sabotage, according to people familiar with the matter. The company says they don’t.

I honestly had no idea that there was a virtual monopoly on Bitcoin mining hardware.

via @bloomberg.com

#anthropic #incidentresponse #aistrategy #securityops #ciso | Michael Mosher I've spent the last week thinking about what the Anthropic AI Agent-based cyberattack should actually change in our security programs. Not the theoretical implications, but the specific moves we need ...

Security recommendations:
www.linkedin.com/posts/michae...

#securitystrategy #aisafety #geopolitics #threatintel | Michael Mosher Thanks to LinkedIn's character limit on posts, I've had to stretch out my thoughts on the AI Agent cyberattack. But this has also given me time to reflect more on its implications... Once the initia...

Broad implications of attack:
www.linkedin.com/posts/michae...

#cybersecurity #threatintel #aitrust #blueteam | Michael Mosher My next installment on the AI Agent cyberattack: Here's the uncomfortable truth about the Anthropic disclosure: we may never get independent verification. The controversy isn't really about whether ...

A look at the reliability of the attack information:
www.linkedin.com/posts/michae...

How GTG-1002 used Claude AI for cyber attack | Michael Mosher posted on the topic | LinkedIn I promised I had more to say on the AI Agent cyber attack. Here's some more analysis on how it played out: What did the GTG-1002 playbook actually look like? Once Anthropic pieced together the tele...

Attack playbook analysis:
www.linkedin.com/posts/michae...

Anthropic's Claude used in Chinese espionage campaign | Michael Mosher posted on the topic | LinkedIn The first clue was speed. Claude Code sessions that looked like routine enterprise testing lit up Anthropic's telemetry in mid-September with thousands of requests per second — far beyond what any hu...

If you would like to jump to one of the posts from here, links below...

What happened:
www.linkedin.com/posts/michae...

Over the past week, I've been posting on LinkedIn about the AI Agent-based cyberattack reported by Anthropic. I've looked at a number of aspects of the attack. The links to all of the posts are summarized here: www.linkedin.com/posts/michae...

#cybersecurity #ai-attack #anthropic #cyberattack

Let’s run the government like a business, drive the car like a bicycle, and play the guitar like a piano

Fancy Bear 'Nearest Neighbor' Attack Uses Nearby Wi-Fi Network

Fancy Bear 'Nearest Neighbor' Attack Uses Nearby Wi-Fi Network

Rising Tides: Wendy Nather on Resilience, Leadership, and Building a Stronger Cybersecurity Community Wendy Nather discusses the importance of amplifying others’ efforts for the common good–and how kindness aids survival.

Rising Tides: Wendy Nather on Resilience, Leadership, and Building a Stronger Cybersecurity Community

👍

After being on Twitter since 2008, it seems like time to set up a new shop.

is there a link to this?

Innovative and frightening at the same time.

We have assumed too long that WiFi attacks would really only be practical through physical proximity.