Instagram is starting to treat an AI-generated person differently from a human creator who happens to use AI
Instagram announced on August 31 that it is renaming its existing 'AI creator' profile disclosure to 'AI-generated profile' and will reduce recommendation reach for accounts centred on synthetic people that fail to disclose that fact. Instagram says properly labelled synthetic profiles will not be penalised simply for using an AI-generated person. The label is not intended for human creators merely using AI to edit photographs, produce graphics, polish captions or make other creative adjustments.
Why visual creators should care
This is a useful distinction for photographers, agencies and AI image creators. A human photographer using AI in production is not being treated as equivalent to an account presenting a person who does not exist. For synthetic-influencer businesses, however, disclosure is moving from an ethical preference to a distribution requirement.
Practical takeaway
If a campaign or account is built around an entirely or substantially synthetic person, record that status in the publishing plan and use the platform’s disclosure tools. Do not assume ordinary AI editing of genuine creator work triggers the same rule.
What to watch next
Watch how Instagram defines borderline hybrid identities, including a real model whose face or body is extensively reconstructed, and whether advertising systems require additional disclosures for synthetic influencers.
Sources
A synthetic war video shows why a plausible image can no longer be treated as evidence of the event it depicts
On August 31 Reuters reported that an AI-generated video was posted alongside a claim that Iran’s Kharg Island was being 'blown to smithereens' during a real escalation between the United States and Iran. Reuters found no evidence that Kharg had been attacked and said its own examination using AI-manipulation detection software indicated that the accompanying footage was most likely synthetically generated. US military activity that had actually taken place concerned Larak Island, hundreds of kilometres away.
Why visual creators should care
For photojournalists, filmmakers and picture desks, this is exactly the environment in which synthetic imagery becomes difficult: a fake visual can be attached to an event that is genuinely unfolding. The factual context makes the image appear more plausible.
Practical takeaway
Do not verify consequential imagery only by asking whether the depicted event sounds plausible. Identify its original source, corroborate the location and event independently, inspect provenance and use forensic tools only as supporting evidence.
What to watch next
Watch how social platforms label synthetic imagery posted by political or official accounts and whether provenance becomes a routine part of newsroom ingestion during fast-moving conflicts.
Sources
Vimeo is drawing its AI disclosure line around whether realistic footage materially changes reality
Vimeo’s current policy requires creators to disclose realistic video that was meaningfully altered or synthetically generated. Examples include showing a real person doing something they did not do, altering footage of an actual event or place, or creating a lifelike scene that never happened. Clearly unrealistic material, ordinary special effects and minor production assistance do not require the same disclosure. Vimeo says labelling AI content does not lower a video’s rating.
Why visual creators should care
This is more useful for working filmmakers than a generic 'AI used' checkbox. It centres disclosure on what the audience might incorrectly believe happened rather than whether machine learning appeared somewhere in post-production.
Practical takeaway
Record whether AI changed the factual meaning of the scene. Routine production assistance and manufacturing a realistic event that never occurred should not be treated as equivalent uses.
What to watch next
Watch how Vimeo handles generative object removal, background replacement and partial reconstruction where the change is visually realistic but less obvious than a full synthetic scene.
Sources
Adobe Stock makes the generator’s licence part of the contributor’s rights check
Adobe Stock’s current generative-AI contributor rules require creators to confirm that the AI tool used permits commercial stock licensing and broad commercial use. Generated material must be labelled during submission. Adobe also requires model releases where generated content depicts, is based on or is intended to portray an identifiable real person, and property releases for recognisable real property. Prompt, title and keyword rules also prohibit references to artist names, known people, fictional characters and protected works without appropriate rights.
Why visual creators should care
Being allowed to download an AI image commercially does not necessarily mean the generator permits the much broader downstream sublicensing involved in stock. Adobe is explicitly making the contributor check that distinction.
Practical takeaway
Before submitting AI imagery to stock, save the generator terms showing that stock licensing is permitted. Keep releases with the generation record where real people or property informed the output.
What to watch next
Watch whether stock agencies begin asking contributors to identify the exact model and version used, rather than relying only on a generic generative-AI declaration.
Sources
Runway says its enterprise protections follow the customer even when the model comes from another provider
Runway’s enterprise documentation says third-party image and video models available inside Runway remain governed by the customer’s enterprise terms and data-processing agreement. Runway says it has contractual commitments that third-party model providers will not train on enterprise customer inputs or outputs. It also says Runway’s enterprise indemnification obligations cover third-party models accessed through the platform.
Why visual creators should care
This is an important contrast with platforms where switching the underlying model can also switch the contractual protection. For agencies working inside Runway Enterprise, the host platform says its enterprise data and indemnity commitments continue across supported third-party models.
Practical takeaway
Record that the model was accessed through the governed Runway Enterprise environment rather than directly through the model provider. The route matters because the contractual protection can depend on where the generation happened.
What to watch next
Watch whether more visual-AI aggregators provide a single contractual wrapper across partner models and whether enterprise interfaces expose that protection clearly at the model-selection stage.
Sources
TikTok’s current US terms separate owning your content from what the platform may learn from it
TikTok’s current US Terms of Service say creators retain ownership of material they create, import or upload but grant the platform a broad licence that includes using content to operate and improve TikTok and to develop technologies, including training, testing and improving machine-learning models and algorithms, subject to applicable settings. For TikTok’s own generative-AI features, users may not present output as human-generated or remove platform-applied watermarks, content-authentication metadata or other disclosure markings.
Why visual creators should care
Ownership and training permission are separate layers. A creator can continue to own a photograph or video while giving a platform contractual rights to process it for technology development.
Practical takeaway
When a client requires no-training treatment, do not rely only on the statement that you retain copyright. Check the platform licence and the relevant account or sharing settings. Preserve any provenance TikTok applies to generated output.
What to watch next
Watch whether TikTok introduces more granular creator controls over machine-learning use and how its US terms interact with contributor, advertiser and business-account agreements.
Sources
