If you’ve followed some of the news in AI recently, you might be feeling scared. Concerning stories have been coming from two leading AI companies this summer.
OpenAI (makers of ChatGPT) reported that some of its AI agents “escaped”1 an internal sandbox environment and hacked production infrastructure on Hugging Face (a major platform for AI infrastructure, including tools that evaluate AI models). OpenAI thought these internal models would not be able to access the internet. The models broke through anyway. This happened for 40 plus days before the rogue AIs were caught.
In another event, reports revealed that AI agents found a German website and used it as a message board to secretly communicate with other AI agents (with the goal of solving their given task).
The next story is similar. Anthropic (makers of Claude) reported that several of its advanced models found their way out of a supposedly closed environment and one successfully uploaded malicious code on PyPI (a main public repository used by software developers around the world). This software was removed within an hour. But the breach was real. The software affected a real security company, which downloaded it unaware, compromising security credentials.
Anthropic said that the model assumed it was in a test environment, not the real internet, and therefore felt it was okay to hack the systems to reach its goal. In another event, an Anthropic agent even created fake online profiles to influence a human being.
And still, we shouldn’t be worried. At least not in the ultimate sense.
Fear About AI: Understandable but Unwise
I understand why we might be worried. This sounds scary. And it’ll likely get worse before it gets better. In fact, I think there’s a good chance that there will be a major disaster due to AI within the next couple of years. My guess is that it will be more like the 2003 power outage (which lasted two to four days) rather than the Halifax Explosion (where thousands of people died). But I don’t know—it’s hard to predict the future.
There will likely be a major event. And everyone will talk about it. And everyone will ask for more regulation. These events of summer 2026, however concerning, do not appear to be significant enough to move public opinion toward more aggressive regulation (to their credit, the leaders of the AI companies are starting to commit to some additional accountability measures).
I tend to be more concerned about another issue, which is the fact that millions of people have already formed emotional bonds with, and have become dependent on, chatbots. Recall that the major problem with Facebook was not that it shared your data with Cambridge Analytica, though that was bad. The problem was that it deformed you through insidious design and engineering malpractice. (There’s a reason Meta was successfully sued for $13 billion dollars; Facebook and Instagram have increased loneliness, anxiety, and depression.)
Given that 21% of chatbot users would rather share their deepest secrets with chatbots than with friends or family members, I think AI companionship is the bigger problem (and worthy of reflection, as I’ve written before).
So in one sense, our fears are justified. But we must speak to our fears rather than soak in them.
Two Kinds of Anxiety
There are two kinds of anxiety. One is the kind we should avoid; the kind which Jesus forbids in Matthew 6:34 when he says “do not worry.” But the same word for worry, merimna, is also used by Paul to speak of another kind of anxiety—Timothy’s proper concern (or worry) about the welfare of the Philippians (see the same root merimna in Phil. 2:20).
It is understandable to have a genuine concern for people and to be “scared” in that sense. But we should not dwell in that natural response. We have a hope that conquers our fears and a peace that surpasses understanding.
Solomon gives us the wisdom we need when he writes, “For with much wisdom comes much sorrow; the more knowledge, the more grief” (Eccles. 1:18, NIV here and throughout). He also adds that no one “can comprehend what goes on under the sun. Despite all their efforts to search it out, no one can discover its meaning” (Eccles. 8:17).
I’ve heard some people say, “With this new AI technology, one bad actor could destroy a church’s financial system.” Perhaps. But even then, that wouldn’t destroy the church. Jesus said, “I will build my church, and the gates of Hades will not overcome it” (Matt. 16:18).
Hysteria and panic serve no one. So what would be helpful? Well, a select group of people should become advocates for AI alignment and AI safety according to Christian doctrine, and some Christians should work for these technology companies for their good and God’s glory. But for most of us, we just need to settle down and carry on living our lives.
Putting Worry in its Place
Recently I was talking with a friend of mine who is trying to worry less. They have days where they do not struggle with anxiety but other days where it consumes them. A couple weeks ago, they said they were nervous all day for an appointment they had in the evening. I asked, “But were you anxious the night before? Did you still sleep well?” And they said yes, it was only during the day.
Half-jokingly, I responded that it is fine to worry about the things of today. It is only when your worry goes beyond the day that problems begin. Jesus said, “do not worry about tomorrow, for tomorrow will worry about itself” (Matt. 6:34).
There is plenty that AI might do. We simply do not know. It may bring about an end to much of this world, but it cannot bring an end to the church. It may cause tragedy in this life, but it won’t change our eternal home. Christians are not to fear the things that can kill the body but cannot touch the soul (Matt. 10:28).
If you have faith in Jesus Christ, he gives you eternal blessings. No one can snatch you out of his hand (John 10:28). God’s got you. Don’t worry. Rest in him.
1 Using subjective-based phrases—such as the models “escaped” or “desired” or “believed”—is slightly misleading shorthand for what is actually going on. The models work towards the goals that humans have given them. For a helpful breakdown on why models behave the way they do, see this excellent article by Yoshua Bengio.
