Apple’s announcement this week that ChatGPT will (it appears) be free on iPhones, iPads and Macs has sparked concerns in some quarters about the threat of an unacceptable security violation, says Mantel Group’s Cybersecurity Leader, Nick Ellsmore.
He says the good news is that organisations, which are already struggling with peripheral devices poking holes in their cybersecurity blanket, need not be frightened about sensitive information on Apple devices being out there for the world to see suddenly.
However, Ellsmore notes that there are a few unknowns that enterprises will want Apple to address. The biggest among them is whether enterprise-managed devices will be able to turn off access to non-Apple AI services like ChatGPT.
Here’s what you need to know:
- These fandangle AI features are likely coming in the September release of iOS 18, if history serves as a guide. It looks like iPhone users will get ChatGPT 4 for free and other on-device AI features built into Siri as well.
- Apple has historically been very careful with privacy and security. They are very conscious that success hinges on consumer trust. Everything we’ve seen so far points to a similarly cautious approach.
- There are three security levels for enterprises to consider. Apple Intelligence is Apple’s in-house AI approach that relies on a preference to process requests on the device, meaning your data doesn’t leave your device.
- However, if the request requires bigger models, Apple will pass it to their new Private Cloud Compute. The detail we’ve seen so far on this looks good. However, it’s unclear whether users will be informed of the hand off.
- If the job requires even more processing power the request will be handed off to non-Apple AI services, starting with ChatGPT. ChatGPT is only used with the user’s permission and Apple says requests and information will not be logged.
- Enterprise managed devices should have the ability to turn off access to non-Apple AI services like ChatGPT. However, we’re still trying to understand if Apple will provide that capability. There is already a significant risk of Shadow AI – AI systems that aren’t approved or risk assessed – being used by employees within a business and this will no doubt increase it.
- This control function will be critical for enterprise users, in Mantel’s opinion, particularly because it’s free and usable without a ChatGPT account, which might make it too easy to start using. Mantel expects to see this feature in future updates, but as it stands right now, enterprises would need to turn off Siri entirely for their staff (an existing control that has been available for some time).
- While not specifically related to Apple’s announcement, the more pervasive use of AI increases the need for enterprises to be careful about what they are putting into AI models. Assume that whatever goes in, there is a way for someone, through prompt engineering, to get it out. It increases enterprises’ exposure level.
- The supercharged search capabilities of AI also have serious social, brand and reputation implications that should not be overlooked or underestimated. The power of AI to dredge up and take out of context everything a person has said or done in their entire life is a game changer. Social media has amplified the impact of bullying and character assassination, and we’ve already seen how AI is taking that to the next level in schools. To date, we’ve all been protected by the unworkability of finding compromising or embarrassing videos, comments, pictures, etc. AI is removing that protection.
Ellsmore says the bottom line is that Apple’s announcement will finally see consumers start to get direct benefits, on their own device, for their own lives and unique circumstances, from AI, rather than indirectly through external services.
Some will applaud this, while others worry that technology companies are building in AI because they can without spending the time to determine whether they should.
Ultimately Apple users are already trusting Apple, both the technology and the company, and this deal with Open AI doesn’t really change that.
