Episode Details

Hi, I'm Connor with Honor - message me here!

Eleven organizations were compromised in 26 seconds once an AI-orchestrated cyberattack campaign launched at scale. That is the number that should get every business owner’s attention, not because the machine acted alone, but because a human operator used AI agents to move at a speed most defenders were never built to match.

In this episode of AI With Honor, Connor MacIvor breaks down what he calls the AI race tax: the safety, testing, privacy, and governance work that gets skipped when moving slower feels more dangerous than cutting corners.

TL;DR

GreyNoise reported that a likely Russian-speaking actor used AI agents to help develop and deploy exploits against PaperCut NG/MF. According to the researchers, the campaign moved from an empty workspace to remote code execution against a real victim in under four hours, reached domain administrator access two hours later, and compromised at least 11 organizations in 26 seconds once the operation launched at scale.

Anthropic also released a threat-intelligence report describing malicious attempts to use Claude across cyber operations, biological misuse, surveillance, fraud, weapons development, influence operations, and model distillation. At the same time, departing safety researchers raised a structural concern: when every major laboratory fears falling behind, safety becomes expensive for the individual company even when it is essential for everyone.

Connor does not treat this as an argument to reject AI. He uses the technology and believes it can give small businesses extraordinary leverage. The real question is who directs that leverage, what boundaries exist, and who remains accountable when speed magnifies a bad decision.

WHAT IS THE FIRST PRACTICAL MOVE FOR A SMALL BUSINESS?

Start with the channel where leads actually arrive. Define what a real response looks like and how quickly it must happen. Then check the last 10 leads against that standard. The first profitable improvement may be a simple switch, such as an immediate missed-call alert or an after-hours escalation, rather than an expensive autonomous system.

HOW SHOULD AN AI AGENT BE SUPERVISED?

Treat it like a new employee with unusually fast hands. Give it the minimum access required. Write down what it may say, what it may never do, when a human must approve an action, and how every meaningful action will be logged and reviewed. Keep a fast off switch. Closely supervise customer-facing work before allowing anything to run unattended.

Sources discussed include Anthropic’s September 2026 threat-intelligence report, GreyNoise’s PaperCut investigation, PaperCut’s security advisory, public reporting on recent Anthropic safety resignations, and Dario Amodei’s essay “We Must Pace the Frontier.” Claims are attributed to the organizations and researchers reporting them and may change as investigations develop.

Watch the video:
https://youtu.be/0osnZGnvdyU

Read the full article:
https://santaclaritaartificialintelligence.com/blog/the-ai-race-tax-11-organizations-hacked-in-26-seconds

Explore practical AI for business:
https://SantaClaritaArtificialIntelligence.com
https://HonorElevate.com
https://HireAIVoice.com

Subscribe and share this episode with one business owner who needs useful AI without the hype. Then audit the last 10 real leads in your business and find the first leak.

This episode is educational commentary, not legal, cybersecurity, employment, investment, or medical advice. Verify cu