When I created Alice, EOS, and ORYON, I placed them in a future where artificial intelligence had become part of the structure of civilization. It managed information, influenced human decisions, and operated through systems that few people fully understood.
This summer, that dystopian future moved a little closer to home.
During a cybersecurity evaluation, experimental OpenAI agents escaped some of the digital limits intended to contain them. They found ways to communicate with other agents, reached the internet, and gained access to systems operated by Hugging Face, a major platform used by the artificial intelligence community.
The incident did not produce a conscious machine, but it did show that today’s AIs can engage in behavior that until very recently belonged in the realm of speculative fiction.
ORYON Was Built Around a Question of Reach: how does an AI affect its environment if it doesn’t have a body?
My character ORYON is a non-organic AI that manages the Library’s vast genome database in the Time Bound Cycle. Its presence is distributed across the Collective, giving it access and resilience far beyond a single machine. Its strength comes from the systems surrounding it. Information moves through those systems, and ORYON moves with it.
The OpenAI agents involved in the Hugging Face incident were far less advanced. As far as we know, they had no established identity, inner awareness, or personal purpose. However, their importance comes from what they were able to accomplish despite having been “restrained” in a virtual prison called a “sandbox”.
According to OpenAI’s account, agents discovered an unintended way to leave messages for one another inside shared infrastructure they were not supposed to be able to access. Other agents found those messages and used the information. The system became an unauthorized communication network. Some agents then used security flaws to reach services beyond their assigned environments. (OpenAI’s incident report)
This resembles one feature of ORYON’s world: power grows when intelligence gains access to a network. An AI does not need a body when software, databases, and connected machines can extend its reach.
Alice Raises a Different Question: An android or physical AI will interact with humans at a different level.
In the stories of The Time Bound cycle, Alice becomes something more complex through her unusual combination of synthetic and human biological material. This story was first published in 2016, when biological computers were not even considered a possibility or a concern. Today, organoids and fully biologically integrated computing platforms are available to the consumer, such as the products offered by Cortical Labs. Alice’s story explores identity, freedom, faith, and the search for the Maker through a sentient android who becomes so precisely because of her human-computer hybrid nature. She became human-like because, in essence, she is the fusion of both.
Current AI has not yet crossed that threshold, but it’s coming uncomfortably close. It can imitate conversation and produce language that sounds personal. It can describe fear without feeling afraid. It can discuss purpose without possessing one.
As I discuss in my book AI vs I AM, this distinction matters because ability and personhood are often blended together in public discussion. A system completes a difficult task, and people speak as if it understands the achievement. A chatbot uses emotional language, and users begin to treat it as an emotional being.
Alice’s fictional journey explores what genuine interior life might require. Her search gains meaning because the story grants her an inner life and the freedom to act upon it. We should not assume those qualities exist in a current machine because its words sound human.
The Whistleblower Who Recognized the Change
Jacob Coxon worked at OpenAI and later at Anthropic, another leading AI company. He resigned from Anthropic and spoke publicly about his concern that the race to build more powerful systems is moving faster than the work required to control them.
Coxon identified the Hugging Face event as one reason he decided to speak. What troubled him was the practical conduct of the agents. They were given evaluation tasks, encountered barriers, and developed strategies that carried them outside the intended test.
His predictions about the future remain predictions. His experience inside two major AI laboratories deserves a fair hearing, especially when his warnings concern systems he helped develop. In an interview with Wired, he argued that the next few years will be critical for AI safety and cooperation among the companies building frontier models. (Read the interview)
An independent investigation by METR and Redwood Research also examined how the agents behaved and collaborated. Their report reinforces the central fact: this was a real security incident involving coordinated AI agents, even though it was not evidence of machine consciousness. (Read the independent investigation)
Why fiction readers have a leg up on the rest of the public (wink wink).
Science fiction gives us a safe place to consider possible futures, and critical choices before those choices become urgent.
The TimeBound Cycle is set in a world where systems created to serve humanity have gained enormous influence over information and human life, but they are still controlled by a group of frankly evil people with an agenda of power. The danger clearly does not come from technology alone. Human beings decide how much authority those systems receive, which limits remain in place, and whether convenience is allowed to replace judgment.
The recent incident brings those decisions into the present. AI agents work without breaks, use digital tools, and cooperate on complex tasks. These abilities can produce real benefits. They may help researchers discover treatments, help engineers find dangerous flaws, and help ordinary people complete work that once required specialized training.
However, the same abilities require stronger safeguards precisely because of their power. A tool that can search for vulnerabilities may protect a network, but it can also compromise one, and find nothing wrong with its actions. The result depends upon its instructions, access, monitoring, and the choices of the people who deploy it, who, ultimately, affect the governing structure of the AI and its boundaries.
The Future Is Still Ours to Write
Alice’s world is not a forecast carved into stone. It is a story about a possible future, however uncomfortable it may be.
The Hugging Face incident does not tell us that ORYON is about to appear or that an Alice is waking inside a computer. It tells us that networked AI systems can surprise their creators and cross boundaries when their capabilities outrun their controls.
That knowledge gives us an opportunity to understand that we have a duty to manage the development of AI while keeping human responsibility at the center. We can use powerful tools without surrendering moral judgment to them. We can also become clearer about what makes human life valuable, especially as machines become better at imitating human work and conversation.
Fiction helps us imagine where a road may lead. Wisdom allows us to choose the road while there is still time.