You wouldn’t hit a man wearing glasses, surely? It is not just the click hungry influencer community or concerned consumers pushing back against ‘pervert glasses’. In fact the debate – and the observations – are as robust on LinkedIn as they are on TikTok. Increasingly, consumer advocates, privacy lawyers, and the academic community are raising questions. And most recently, the Federal Government. Attorney-General, Michelle Rowland's intervention in August addressed the wider market, and not Meta specifically. She noted smart glasses may have significant privacy implications, especially for women and children who could be disproportionately affected by inappropriate recording, harassment, surveillance or other harmful conduct enabled by this technology. “Unlike other forms of technology, smart glasses may be used more discreetly, making it harder to know when you are being recorded … The Government is continuing to work on the next phase of privacy reforms to ensure our privacy laws are fit for purpose in the digital age.” Ancient grudge, new mutiny The next phase of those reforms is now taking shape, and smart glasses are explicitly in the government's sights. According to Rowland, "The reforms will bolster privacy protections for Australians and help tackle emerging risks from new technologies including Artificial Intelligence, and wearable devices such as smart glasses." An Attorney-General's Department consultation paper on the proposed Privacy Amendment (Personal Data Protection) Bill 2026 identifies wearable surveillance technologies, including smart glasses and earbuds, among the emerging technologies the Government wants the next round of privacy reforms to address.  Under the exposure draft, existing Australian Privacy Principles governing collection, unsolicited information and use and disclosure would be reorganised around a broader requirement that collection, use and disclosure of personal information be lawful and “fair and reasonable in the circumstances”.  That assessment would require companies to consider what a reasonable person would expect, whether the company has been transparent, whether less personal information could achieve the same purpose, whether the person had a genuine choice and whether the privacy impact and potential harm are proportionate to the benefits. Where the information relates to a child, the child's best interests would become a primary consideration.  The changes would also broaden the definition of personal information so a person's legal identity need not be known if they can be recognised or singled out, while information generated or derived through technological processes can constitute collection.   The Government's consultation paper links those changes to smart glasses in its discussion of wearable surveillance technologies. The paper says information collected through the devices can include photos, video, audio and AI-generated inferences, and says the reforms are intended to ensure that inferences drawn by AI-enabled technologies are treated as collection under the Privacy Act.  Among the questions it is putting to stakeholders is: “To what extent can the reforms ensure that consent is meaningful, given that wearable technologies compromise the transparency of captures and uses of individuals’ personal information?”  The exposure draft, of course, remains subject to consultation and could change before legislation is introduced or enacted. But that's for the future. Meta will find complications enough in the current laws. Take the issue of consent. While Meta can provide information to the person wearing the glasses during setup, it cannot ensure that people standing nearby receive an adequate privacy notice. For instance, a recording light may alert a bystander that a device is operating, but it does not explain that the person’s face, voice or surroundings could be collected, transcribed, retained, reviewed, disclosed overseas or used to improve artificial intelligence systems. That’s a distinction could bring the glasses into conflict with Australian Privacy Principle 5, which requires organisations to take reasonable steps to notify people about the collection of their personal information or otherwise ensure they are aware of it. The adequacy of notice is likely to depend partly on context. A visible recording indicator may carry more meaning when someone is deliberately filming a public event than when glasses activate an AI feature during an ordinary conversation. Likewise, a wearer’s acceptance of Meta’s conditions cannot bind another person. Where Australian law requires consent, the wearer cannot provide it on behalf of a bystander. Nor would the wearer’s agreement necessarily discharge Meta’s own responsibilities for collecting information, notifying the person concerned and controlling how the information is subsequently used or disclosed. Those questions potentially engage Australian Privacy Principles 3, 5 and 6. Civic Data's Brinkworth says the public debate is being argued in the wrong register. “Everyone reaches for consent, and for a bystander consent is a fiction. The obligation that actually bites under the Privacy Act is notification. APP 5 asks what steps, if any, are reasonable to make a person aware their information is being collected, and the Bunnings decision is the warning here: The Tribunal excused the collection itself, but it left the finding that customers were not properly notified standing, and the Commissioner did not appeal. A blinking light tells you a device is on. It does not tell you your face and voice may be transcribed, reviewed by a person, sent offshore or used to improve a model.” Consent misconceptions For Macquarie University cybersecurity and privacy researcher, Dali Kaafar, the privacy problem created by Meta’s smart glasses begins with a basic misunderstanding about what the devices are capable of doing. Consumers may see a pair of glasses with a camera attached, he said, but the technology increasingly acts as an input pipeline for artificial intelligence systems, capturing information from the physical world that can be interpreted, searched and combined with other data. “It’s not a normal camera. It’s not just capturing images,” Kaafar says. “They capture and increasingly interpret everything in the view in the scene of these cameras, and that feeds pretty much into AI systems we cannot see.” For his part, Brinkworth says the Act draws its line not at the camera but at what the system does with the capture; biometric identification is where consent becomes mandatory, which is why a facial-recognition feature would change everything. Kaafar said Meta leaves the consent model attached to the person operating the technology rather than many of the people whose information can be captured. “The consent Meta and all the other providers are putting in their terms of use is sitting with the wrong person,” he said. “It’s not about who wears these devices; it’s about the people who are just in the frame, who obviously never agreed to anything like that.” The concern extends beyond the initial recording. Kaafar said images, voices and other contextual information can become inputs into data systems whose downstream processing may be difficult for consumers to trace or control. “That’s data that we can’t see, we can’t trace, we can’t even control,” he said. Kaafar says that also creates a cybersecurity problem that differs from conventional data breaches, where companies and regulators can often quantify the damage by counting compromised customer records. Smart-glasses data could contain images, voices or contextual information involving large numbers of people who never had accounts with the company and whose identities may not have been recorded alongside the underlying data. And that, it turn, could make it difficult to determine how many people had been affected if a large dataset were compromised. “We’re in a position that I think has never been seen before and never been really assessed before, where data can be pretty much every single individual, every single citizen,” Kaafar says. For Kaafar, conventional privacy frameworks also tend to focus heavily on whether information directly identifies a person or can be used to re-identify them. AI systems complicate that distinction because faces and other pieces of information can act as what researchers call quasi-identifiers, allowing additional characteristics or identities to be inferred after the data is combined with other sources. “If I don’t know your name and I only see your face, or part of your face, is that already a quasi-identifier that will allow me to kind of infer all sorts of things about you,” he says. Kaafar expects regulatory challenge to grow as AI-equipped glasses became cheaper and more widely available from manufacturers beyond large technology companies such as Meta. Policymakers have as yet had a limited opportunity to establish appropriate safeguards before widespread adoption made the technology part of everyday life. “The pattern we keep repeating is ‘technologies are made desirable first, regulated after,’” he says.