Wild watching so many businesses race to deploy AI as the answer to everything. Open-weight models and no-guardrails frontier systems wired into agents doing who knows what.
We heard all about it with ransomware too, right up until payroll stopped running, the water company couldn’t invoice, supply chains broke, and hospitals had to drop to paper. It’s all hype, security folks always cry wolf, they’re the department of no, and they slow us down.
The government pulled Mythos/Fable off the market over a jailbreak, and Anthropic’s own answer was that the same capabilities were essentially already available on other public models. Yep. Their reassurance to us is that the capability is already everywhere. GPT 5.5 isn’t far behind, OpenRouter’s Fusion is right there at a tenth the cost, Microsoft’s MDASH outperformed Mythos Preview, and the leapfrogging will continue.
Back in late November, when Opus 4.5 landed, many felt that shift. It was a radical jump in red teaming and vulnerability research, and that was generations ago… seven months. The speed at which I could surface critical vulns in production systems was eye-opening, at least to those in the industry, while many outside simply attributed it to the usual parlor trick.
Mythos/Fable isn’t a new category of risk. It’s the same risk, scaled and handed to anyone with an API key. Their own, their company’s, or one they stole. What used to take a skilled operator now takes simple access. And the national security angle of pausing the model isn’t what’s going to protect businesses.
You can’t speed run adoption while you crawl on defense. These tools are already in the hands of the people breaking in, and the pressure behind them is building fast. There’s no degraded mode where you limp along on half a team. You either have people on the wall or you don’t. And as budgets pour into tokens, security will be one of those lines that’s squeezed to pay for it.
The engineers and analysts and threat hunters who secure what everyone else is so eager to ship belong at the tip of the spear, not on the cleanup crew after it breaks. In front of what we build, not behind it. That’s the investment that decides whether the next wake up call is a headline, defensible cyber claim, or another notch on the casualty list.