Oh, Max!

September 30, 2026

Matthew Frazer

There is a trope of popular fiction where the (typically male) flawed hero blunders through the adventure with either brute strength or blind confidence as his chief resource, and his success is assured through the assistance of an intelligent, long-suffering (typically female) companion.

Think of Agent 99 in the classic TV series Get Smart. Or Penny from Inspector Gadget. Or Miss Moneypenny in (some versions of) James Bond.

It can make for some comical set-ups, leaning into the classic gender stereotypes: men are brash, over-confident, people of action who get all the glory; women are quiet, competent, intelligent supporters who either shun the limelight or never get the recognition they deserve.

In the modern incarnation of this fictional trope, the ‘assistant’ is not necessarily human. The video game series Halo has Cortana, a highly competent digital strategic assistant who complements the military strength of Master Chief with intelligence, hacking skills and sarcastic wit. In the later Avengers movies, ‘Friday’ is a female-sounding AI voice in Tony Stark’s ear, helping him out of many tough spots. (His earlier ‘male’ AI assistant unfortunately gave rise to a megalomaniacal robot, Ultron.)

As we enter the age of ubiquitous AI in the non-fictional world, we encounter an interesting phenomenon. We regularly find ourselves interacting with disembodied ‘persons’. For reasons of user comfort and/or familiarity these are often given a persona that includes gender attributes. 

Many of the early entries into the virtual assistant space were given feminine characteristics. Siri, arguably the most famous mass-market voice assistant, was predominantly conceived of as ‘a she’.[1] Cortana, one of Microsoft’s more recent attempts, was also female—adapted from the Halo video game franchise mentioned above (though the real-world offering was far inferior to the fictional one).

Initially, these assistants were a neat party trick,[2] able to generate human-looking text and blitzing the venerable Turing Test for determining whether an algorithm constituted artificial intelligence or not. But in only a few short years they have become a piece of critical social infrastructure as well as having significant economic roles. They can locate and summarise information, synthesise solutions to problems, show ‘creativity’, write code, music, poetry and more.

From even their early, pre-ChatGPT incarnations, the issue of problematic ‘feminised’ behaviour from these assistants was identified. In 2019, UNESCO[3] catalogued the kinds of responses that female-presenting voice assistants gave to sexualised or abusive prompts—in most cases either passive or tolerant, yet in many ways playful or even ‘flirty’. Researchers at Quartz found that these kinds of responses ‘reinforce stereotypes of unassertive, subservient women in service positions … [and] intensify rape culture by presenting indirect ambiguity as a valid response to harassment’. [4]

Intriguingly, or perhaps alarmingly, AI ‘fembots’ have also found a breakout role in providing companionship. There are now entire companies whose business model is providing ‘AI friends’ for paying customers. These friends can have their own distinctive personalities and never tire, move away, or argue. Some people think of them like a partner or spouse, sharing updates on daily life, confiding in them, seeking counsel, and engaging in intimate or explicit verbal interactions.

This is not a merely hypothetical discussion around how virtual assistants may respond to unhealthy or abusive conversations. It is really happening, frighteningly prevalent, and growing in popularity. A 2021 study found that about 20.4% of conversations sampled with digital assistants contained some form of abusive language, and that systems that presented with a female voice received substantially more sexual harassment and sexist abuse than those with no specific gender markers. [5]

The Australian Government’s eSafety Commissioner found that, in 2026, around 78% of Australian children (aged 10-17) had used an AI Assistant, most of them on a regular basis. About half had used the service for personal reasons such as physical or mental health advice or companionship, and 20% had experienced potentially inappropriate or harmful interactions with these tools.[6]

Large parts of society, then, including children, are interacting with digital assistants. Many of these assistants personify feminine stereotypes and provide no resistance or push-back when presented with problematic or abusive interactions. Casting a subservient, endlessly nurturing, sycophantic digital ‘slave’ as a female character does not bode well for breaking down some of the more harmful concepts of masculinity we find in the world today.

Large Language Models are trained largely on ‘real’ human text, and so what we witness is, in part, a reflection of what society already is. But we can, and should, consider carefully how we might develop these systems to improve what society may become.

We may question whether there is a moral responsibility to treat Artificial Intelligence systems with kindness given they are not, in fact, human but piles of probability distributions stuffed inside a trench coat (like Maxwell Smart’s?). This is a fair conundrum, but the more pressing question is whether this kind of behaviour leaks into human relationships where the moral imperative is much clearer.

There is evidence that such ‘leakage’ does occur. Weisman and Janardhanan found that managers who interacted with feminised Voice Assistants changed the way they interacted with their subordinate female workers. They expected more immediate help from female than male subordinates, penalised them more strongly when help was not forthcoming, and spoke to them more impersonally. The same result was not not observed with male equivalents.[7]

Christians must guard against the devil and take great care where we allow our hearts to dwell. Jesus said:

For the mouth speaks out of that which fills the heart. The good man brings out of his good treasure what is good; and the evil man brings out of his evil treasure what is evil.  But I tell you that every careless word that people speak, they shall give an accounting for it in the day of judgment. (Matthew 12:34b-36, NASB)

Our private thought life—and by extension our private chat history—serves as rehearsal for our outward interactions. And while nobody is suggesting we should send Siri a Christmas card every year, we, whoever we are, should be mindful of how our own attitudes towards human women might be affected, even subtly, by regularly interacting with ‘female assistants’ who are docile, compliant, and endlessly tolerant of all kinds of abuse.

The Bible does not promote or condone a masculinity that uses its power to dominate women (or others) as inferiors. Rather, it encourages men to employ the power and strength they have for the good of the women (and others) in their orbit—even to the point of self-sacrifice, as Christ did for the church. Regular interactions with non-human feminised personalities can give the male user an illusion of power with no responsibility: no imperative for mutual respect; no accountability; no need to compromise. This is a dangerous diet to feed a man’s heart, especially when he is already subject to society’s diverse and somewhat chaotic notions of appropriate relationships between men and women.

Our AI creations are useful tools but they have the capacity to exploit moral fault lines. Christians must exercise wisdom when using these technologies and think carefully about how we frame our interactions with these pseudo-people to avoid warping our patterns of relating to real people, or amplifying harmful predilections. You may feel silly saying ‘please’ to Siri, but it’s a small price to pay to avoid becoming habituated to making uncivil demands of your spouse or employees.

As copy editors, search engines or vibe-coding tools, AI tools may be perfectly adequate.[8] As surrogates for healthy human relationships, they have ‘missed it by that much!’

Leave a comment

Comments will be approved before showing up.

Name *

Email *

Comment *

---

Also in CASE Online Library