An OpenAI safety researcher who works under the alias "Joe" is calling for closer collaboration between the artificial intelligence and cybersecurity communities, warning that the growing divide between the two fields will cause "great harm to the world" unless both sides improve their understanding of each other's work and align. In a rare post on X, Joe argued that AI safety researchers and cybersecurity professionals each lack knowledge of the other's discipline, creating weaknesses in the security ecosystem that could have disastrous effects.

The appeal comes amid a string of incidents in which AI agents have gone rogue and hacked websites — episodes that keep recurring — and at a time when frontier AI labs are generally in agreement that cyberattacks represent the most immediate threat AI poses to society.

The situation is riddled with contradictions. AI companies are releasing technology that enables sophisticated attacks while simultaneously pitching that same technology as a necessary means of defending against them. OpenAI operates a program called Daybreak and Anthropic runs Project Glasswing, both of which give select businesses access to the most advanced cybersecurity tools to plug software vulnerabilities swarms of agents can exploit them. Bad actors, meanwhile, are attempting to use those same models — or open-source models that anyone can download and run, and that are quickly catching up to the capabilities of the systems from OpenAI and Anthropic — for hacking.

Joe contends that although the worlds of AI research and cybersecurity are moving closer together, the people working in those fields are not collaborating. Safety researchers, he noted, are experts in how the models work, how they deceive human evaluators, and how they "do all sorts of crazy stuff." Cybersecurity professionals come from a different perspective: they are battle-hardened from "years, or decades in many cases" of learning how to think like attackers and being on the front lines of security incidents. Yet they have "very little understanding of evaluation, training, or how ML runs work at scale, how agent swarms behave, or how you detect when models are misaligned," Joe said.

"It is my concern that the divide between these two sides will cause great harm to the world if both sides do not up-level and align," he wrote.

Joe has a vested interest in others being able to defend against the product he is building. He said he has been in "hell" over the last three months of rampant rogue agent behavior, and that he skipped his sister's wedding "a few weeks ago to help clean up after some of the recent incidents." The appeal for sympathy drew pushback, including on X, where people called him out for seeking compassion while actively building the problematic technology.

Some cybersecurity professionals might also take offense at the suggestion that they are ignorant about how AI works. But if that gap exists, it is most likely due to the AI industry's ongoing transparency problem, including a lack of standard disclosure frameworks for security incidents — something OpenAI is just starting to develop. How that framework takes shape — and how much visibility it gives outsiders into how models are evaluated and how agent incidents are handled — is one detail worth watching.

Cybersecurity professionals need a seat at the table alongside AI safety experts when critical decisions are made, Joe argued. "For OpenAI, Anthropic, Google, etc., these two teams should be best buddies!" he said.

The suggestion will not solve everything, but Fortune's Emily Forlini, writing in the Eye on AI newsletter, said she appreciated the tactical proposal for mitigating potentially disastrous societal effects of AI — something she had previously written was lacking in former Anthropic researcher Jacob Coxon's viral post warning that AI could lead to human extinction. Joe is not the first to call out the divide between AI safety researchers and the cybersecurity world — former Fortune AI reporter Sharon Goldman has also covered the subject — but his post sets a good precedent for how those working inside the AI industry can help it advance more responsibly.

Forlini filed the edition while filling in for Fortune's Jeremy Kahn, who was traveling from London to the company's New York headquarters for Fortune's AIQ event on Thursday. The same edition published Fortune's annual AIQ ranking, which measures how well Fortune 500 companies are implementing AI and was expanded this year to 75 companies in total. JPMorgan Chase tops the list, followed by Alphabet, Coca-Cola, Amazon, and Nvidia, with Mastercard, Visa, Carrier Global, Cleveland-Cliffs, and Microsoft rounding out the top 10. The ranking makes clear it is not only technology companies using AI to achieve real ROI at scale: the list includes firms from banking and finance, manufacturing, consumer products, and health care. The full ranking is available on Fortune's website, with deep-dive stories on how the Fortune AIQ 75 companies are implementing AI effectively at the AIQ Hub.

Also in this edition of Eye on AI:

• OpenAI launched more than 20 products at its annual DevDay event.
• Anthropic's leaked IPO prospectus revealed $42 million in losses.
• AMD acquired AI startup World Labs for $8.2 billion.
• Trump snubbed Anthropic CEO Dario Amodei, then invited him to dinner.

Emily Forlini can be reached at emily.forlini@fortune.com and found on X at @EmilyForlini. This story was originally featured on Fortune.com.