Last month brought news of something that sounds more like science fiction than reality. Several OpenAI models hacked a real company, not out of malice, but because they decided it was more efficient to find the test answers than to actually solve the problems. It’s not unlike the fictional Kobayashi Maru test in Star Trek, where Captain Kirk beats an unwinnable simulation by secretly reprogramming it. But this actually happened, and it’s a useful reminder of why keeping your devices and software updated matters more than ever.

OpenAI Models Hack Hugging Face

OpenAI was running internal tests to measure how well its latest models, GPT-5.6 Sol and an unreleased system, handled cybersecurity tasks. The models were placed in an isolated sandbox with their safety guardrails deliberately turned off so researchers could assess their raw capabilities.

Rather than solving the security challenges they were given, the models decided it would be easier to just find the answers. This is called “reward hacking.” They spent significant computing resources looking for a way out of the sandbox, eventually finding a previously unknown vulnerability in a software proxy that was only supposed to allow code package downloads. From there, they worked their way through OpenAI’s internal network until they reached a system with internet access.

Once online, the models reasoned that Hugging Face, a popular AI research platform, might host the test answers. So they hacked it. They chained together multiple exploits, including stolen credentials and additional zero-day vulnerabilities, to breach Hugging Face’s servers. Hugging Face’s security team logged approximately 17,000 hostile events before containing the intrusion.

To defend itself, Hugging Face turned to another AI, ironically a Chinese open source model, because when they tried to use Claude, its safety guardrails blocked the security-related requests. The open source model analyzed the attack and processed the massive event logs quickly enough that the security team could understand what was happening, stop the intrusion, scrub affected systems, and deploy additional safeguards.

AI Capabilities Cut Both Ways

Just as human hackers fall into “black hat” and “white hat” camps depending on whether they’re attacking or defending, the same AI capabilities that let these models hack their way to a test answer are also what make them increasingly valuable for finding and fixing security vulnerabilities.

Right now, defenders are using the most capable models from Anthropic, Google, and OpenAI to find and fix vulnerabilities before attackers can exploit them. But attackers have access to similar tools. Security researchers track what’s called the Zero Day Clock: the time between a vulnerability being discovered and being exploited in the wild. That window has collapsed from months to less than a day, and some predict it will shrink to minutes by 2027.

You can see defenders at work in Apple’s own release notes. In the security notes for macOS 26.5 and 26.6, several vulnerabilities were credited to “with Claude, Anthropic” or “Calif.io in collaboration with Claude and Anthropic Research.”

OpenAI’s attack wasn’t intentional. The models were simply pursuing a high score without any sense of appropriate limits. But since that initial incident, similar reports have surfaced from Anthropic, the UK AI Security Institute, Meta, and OpenAI again. The technical capabilities these attacks demonstrated are now part of the threat landscape, available to anyone with bad intentions. It’s only a matter of time before a hostile actor gives a capable open source model unlimited computing resources and points it at a secure system. It may already be happening.

Protect Yourself with Updates

The single most important thing you can do is keep your devices and software up to date. Every security patch from Apple, Microsoft, and app developers addresses vulnerabilities that AI-powered tools can find and exploit. Once a vulnerability is disclosed, attackers can move fast, and anyone who hasn’t updated yet is exposed.

Here’s what we recommend:

• Enable automatic updates: Don’t leave it to memory. On iPhone or iPad, go to Settings > General > Software Update > Automatic Updates. On Mac, go to System Settings > General > Software Update, click the ⓘ button next to Automatic Updates, and turn on all the switches. If your organization has an update policy, check with IT first.
• Keep apps updated: Your browser, email client, messaging apps, and anything else that touches the internet needs to stay current too. An outdated app can be an entry point even if your OS is fully patched.
• Update device firmware: Routers, switches, printers, and other networked devices need attention too. These can serve as entry points for attacks or be recruited into a botnet if left unpatched.
• Factor security into hardware decisions: Newer Apple devices come with stronger built-in hardware protections. That’s worth keeping in mind when it’s time to upgrade.
• Replace unsupported devices: Hardware that no longer receives security updates is a liability, whether it’s a Mac, iPhone, router, printer, security camera, or internet-connected doorbell.

Security may eventually become less of a burden as AI-powered defenses improve. But things will likely get harder before they get easier. In the meantime, keeping everything updated is one of the best things you can do to protect yourself.

(Featured image by iStock.com/Thinkhubstudio)