20 August 2024
Words – Jacky Winter
Illustration – Sebastian Cestaro

Whether you think Meta's new AI policy is a hot mess of "nothing to see here" or gaslighty tech optimism, our friends from Jacky Winter are here to sort fact from fiction with a deep dive into what it all means for Instagram users, and what you can do about it.

Over the past few months we've been fielding a lot of questions from artists regarding Meta's announcement that from 26 June 2024, it intends to scrape images from its platforms (including Instagram) to train future AI models. Most creatives, ourselves included, are trying to weigh up the benefits of continuing to post work on these platforms vs the risk of having the work appropriated for potential commercial use without compensation.

To help address these questions, we've put together a short primer of where things are at based on our own internal readings of the terms, statements and articles available on the topic, in hopes that it might offer some clarity. The AI landscape is constantly evolving and this isn’t legal advice or anything, so please take everything with a grain of salt - it's simply our best attempt at trying to make sense of things, where they’re sitting right now.

WHAT HAPPENS TO CONTENT THAT HAS BEEN UPLOADED TO META'S PLATFORMS?

• Anything uploaded can be used to train their AI models, and it’s possible the scraping began some time ago. Meta has stated that it had already started training AI on Facebook and Instagram posts, prior to 26 June. There are reports that content dating as far back as 2007 has been scraped.
  
  
• The 26 June 2024 date when the scraping was due to start was actually relevant only to the EU/UK. Due to the privacy and data laws there, Meta had to tell people in advance and give them a way to opt out. In other parts of the world (including Australia and America) these laws don't apply, so Meta had already proceeded with the data scraping.
  
  
  

  ---

    

CAN I OPT OUT?

Opting out of Meta’s AI training is available to users in the EU/UK. As far as we’re aware, there is currently no opt-out option for users located in other countries.

• For those in the EU/UK looking to opt out, a quick Google search will bring up many hits outlining how to do it - such as this article by MIT technology Review which covers both Facebook and Instagram.
  
  
• It should be noted that while artists in the EU/UK can opt out from their accounts, this doesn't extend to work that has been shared by third parties (such as agents, brands, clients, collaborators, studios and other media organisations) who have not or are unable to opt out. So be mindful that any work that has been shared or reposted to another account could still be scraped, even if you have personally opted out.
  
  
• You might have seen people posting "I own copyright and don’t consent to Meta using my content" type statements to their stories. These don't work and don't override the incredibly broad licence you grant Meta when you agree to the terms and conditions required to use its platforms.
  
  

  ---

    

DOES DELETING POSTS HELP?

• Deleting your posts will stop future training or re-training, but if the content has already been scraped, deleting will have no effect. In other words, deleting your post will not remove your content from any AI model that has already scraped it.
  
  
• As an alternative to deleting posts, changing your profile to private might help. Meta has said that it doesn’t train AI models on posts from private Facebook and Instagram accounts. However, the license granted in Meta’s user terms and conditions is technically wide enough to cover private content, so Meta could decide to use content from private accounts down the track if they wanted to.
  
  

HOW CAN I PROTECT MY WORK FROM AI SCRAPING?

Check out our detailed article which covers the various tools you can use to protect your work from being used to train generative AI models.

WHAT ARE META GOING TO DO WITH THE DATA?

Meta have been very vague and broad about this. All we know is that it will be used to train and improve its AI tools, which currently include image and avatar generators, and are likely to expand to other areas and applications in future.

WHERE DO WE GO IF WE DON’T HAVE INSTAGRAM?

• Cara has gotten a lot of attention recently for its integration of anti-AI software Glaze, and it’s overall anti-AI stance. The main question here is whether there will be enough people actually commissioning art on there to make it worthwhile.
  
  
• Anything that is publicly viewable on the internet can be scraped - if not by Meta, then potentially by other AI platforms. For example, Getty Images is currently suing Stablility AI which was trained on watermarked Getty images and thumbnails which can be viewed by anyone. So if the goal is to limit scraping, apart from going offline completely, the safest option is to go private. This may mean implementing a paywall, or relying on something more old-school like email - although this does limit your reach to those subscribed to your mailing list or already in your network.
  
  
• It wouldn't be a Jacky Winter post if we didn't plug our own newsletter app, so if you are interested in going deeper into email newsletters - do check out PencilBooth.com!
  
  
• Outside of moving work to an anti-AI / private / non-Meta platform, there are also more technical methods available (including artwork cloaking software such as Glaze and Nighshade, and do-not-train web tools via Spawning.ai) which can be used to protect individual artworks and websites from being scraped.
  
  
  

  ---

    

SHOULDN’T THIS KIND OF THING BE ILLEGAL?

AI has been progressing at breakneck pace and laws and regulations are playing catch up.

At the time of writing, only the EU has laws on the regulation of AI. The legislation, known as the Artificial Intelligence Act, only just entered into force on 1 August 2024 (and isn’t scheduled to be fully implemented until July 2026) so we will be waiting to see what effects the provisions have on companies such as Meta and whether other countries will follow suit.

For those interested, the EU legislation follows a tiered, risk-based approach, which means regulations are stricter for AI uses with a higher risk of social harm (for example in areas such as public safety, security, and fundamental human rights).

Generative AI art falls within the “general-purpose AI” category, which is subject mainly to transparency requirements:

• People must be told they are interacting with AI generated content (especially applicable to deep fakes)
  
  
• AI companies will need to provide summaries of the copyright data used to train their models
  
  
• AI models will also be required to have safeguards against the generation of content that breaches EU laws
  
  

The latter two requirements appear to be in place as a protection against IP infringement. However, we still don't know the specific circumstances in which data scraping or use of AI generated material would be considered illegal from a infringement standpoint. It is likely that we will need to wait until these clauses are interpreted in courts for further clarity – but, it’s a start.

SO WHAT NOW?

Overall, AI remains a very messy area and it may be some time before we make it out of this AI haze. If you’re in need of a bit of copium in the meantime, here are some possible solutions and hopeful outcomes that have been thrown around:

• Clever people, such as those at The Glaze Project, Spawning.ai and overl.ai have been developing tools to deter AI bots or protect images from being scraped. We’ve written an article on how to protect your art from AI here [LINK].
  
  
• A key hope is that AI models trained on content that wasn't expressly licensed for that purpose will eventually be deemed unlawful in most countries and die out (much like Napster, Limewire and Pirate Bay did). After that, "legit" opt-in databases will take their place. We think a few of these exist already, but basically these would be platforms where artists submit their own work for the purpose of AI training and get paid a license fee or royalties in return.
  
  
• Apparently there isn't enough electricity in the world to enable AI to get to the "next stage". So for all the crazy advances we've seen in the last few months, maybe it will be forced to slow down after this just because there isn't enough computing power for it. Food for thought!
  
  

So where does that leave us at the end of the day? Personally we still feel the rewards of maintaining our presence on Instagram outweigh the potential risks, but it is something we are closely monitoring and if and when there's a more substantial update to post, we'll do so via our channels.

Guest post by Jacky Winter, a next-gen artist agency that gives creative ideas and careers a place to soar. 

Website  |  Instagram  |  PencilBooth  |  LinkedIn