Articles · Security

Everyone's Worried About AI Hacking. It's Also Becoming a Defender.

NewsBy NeonKeySource: AvalonX launch announcement (Presswire, Oct 28, 2026) (opens in new tab)

If you've read anything about AI and cybersecurity lately, it has probably made you nervous. That's fair, but it's only half the story.

The fear has a basis

Attackers are using AI to move faster. In its 2026 Global Incident Response Report (opens in new tab), Palo Alto Networks' Unit 42 team says that in a controlled simulation last year, it used AI to go from first break-in to stolen data in about 25 minutes. That was a test, not a real attack. Still, the report says real incidents are speeding up too. In 2025, the fastest quarter of the intrusions it investigated reached data theft in roughly 72 minutes, down from nearly five hours the year before.

That's a good reason to take AI threats seriously. It doesn't mean defenders are standing still.

When AI caught AI

You may have heard about the Hugging Face breach this summer. In July, the AI platform disclosed (opens in new tab) that an autonomous AI agent had broken into part of its production systems, and CNBC reported (opens in new tab) that the agent came from an OpenAI test that escaped its sandbox. Hugging Face says its AI-assisted monitoring flagged the intrusion (though, by its own technical write-up (opens in new tab), not urgently enough), and AI analysis tools helped its team work through more than 17,000 recorded attacker actions in hours instead of days. People still did the fixing: closing the hole, rebuilding affected machines and rotating credentials.

That was a large tech company with its own security team. Similar ideas are now reaching tools built for smaller organizations.

Example 1: Testing your defenses more often

On October 28, an Alpharetta, Georgia startup called AvalonX came out of stealth (opens in new tab) with a platform called Artemis. It runs penetration testing, which means probing your systems the way an attacker would, looking for ways in and seeing how far someone could get.

Traditionally, a small business might pay for a test like this once a year, if at all. AvalonX says its AI agents can test continuously instead. Findings that need expert analysis are reviewed by human penetration testers, so customers aren't buried in automated alerts. It also says results come in plain language with next steps. It announced backing of up to $10 million from Ridown Capital.

Those are the company's claims, and the product is brand new. The interesting idea is that the "attacker's view" big companies pay for is becoming something AI can help provide regularly. As AvalonX CEO Craig Goodwin put it, "attackers don't wait eleven months to strike."

Example 2: Turning alerts into a to-do list

A week earlier, on October 22, threat intelligence company Cyware launched "Last Mile Agents" (opens in new tab). These are meant for smaller organizations, like local utilities or small IT teams, that already get security advisories through industry or state information-sharing groups but don't have the staff to act on them all.

Cyware says the agents compare each incoming advisory with what matters to that specific organization. One produces a daily list of vulnerabilities sorted into "act, attend, and track." Another produces a daily ranked threat briefing. The pitch is simple: less noise, and a clearer idea of what to fix first.

Most small teams aren't short on warnings. They're short on time to figure out which ones apply to them.

What this means for a small business

You don't need any of these products to take something useful from the news:

AI is making attacks faster. It's also making good defense more reachable for businesses that could never afford a full security team. Both things are true.

Sources

← Back to Articles