AI/Nov 10 edition/Reporting & analysis
PolicyMultimodalResearch

PolicyLaw, regulation & governance

Photography schools add AI modules and usage rules as Austrian students reject Midjourney as a main way to make images

Photography schools are adding AI training and usage rules, such as Spéos's ten-session module and CCS's documentation requirements. Meanwhile, most of 39 Digital Arts students surveyed in Austria rejected Midjourney as a main way to make images, saying photography depends on engaging with the real world.

From the Financial Times · AI article of November 10, 2026 and cited sources: In an EDULEARN26 study from Upper Austria, students took conventional camera, lighting and post-production training and then used Midjourney. [2]

SOURCES[2] IATED: Teaching Photography in the Age of Generative AI… library.iated.org

AI-assisted, not human-edited. Methodology Corrections Who owns this site

KEY FACTS
What happened
In an EDULEARN26 study from Upper Austria, students took conventional camera, lighting and post-production training and then used Midjourney. [2]
Original source
Financial Times · AI article, independent perspective
Brief by
AT Pulse research desk AI-assisted; publisher and contact on the desk page
When
November 10, 2026 article published
Evidence
7 sources across 7 sites

Reach: not measured yet

THE CORE IDEAS3 TAKEAWAYS
01

In an EDULEARN26 study from Upper Austria, students took conventional camera, lighting and post-production training and then used Midjourney. Afterwards, most saw prompting and photography as fundamentally different practices. They accepted AI for auxiliary tasks such as masking and early ideation. The study covered one cohort with no control group, so it measures perceptions rather than skills. [2]

02

Schools are taking different approaches. Spéos will run ten AI sessions from September 2026, covering retouching, image generation and workflow automation. CCS lets photography students use Adobe's built-in AI for small edits without citing it, but any other AI tool needs written instructor approval, and students must keep unedited originals. A Gothenburg exercise has students alter real photos with generative tools over 180 minutes and then critique the results. [1] [5] [6] [7]

03

Outside the classroom, rules are stricter. In 2023, Boris Eldagsen refused a Sony World Photography Award for an AI-generated entry. Later that year, World Press Photo banned generative fill and fully generated images in all categories, citing accuracy and trust. CCS, by contrast, allows built-in AI edits without citation. [3] [4] [7]

WHY IT MATTERS

The evidence is one 39-person survey, one article describing a teaching method, and school policies.

Read the full assessment

The likely result is graduates trained in AI retouching but wary of AI as a way to author images, while editorial buyers will need provenance controls that match contest bans.

Executive brief

Students may be more sceptical of generative imagery than the schools teaching them. In a survey of 39 Digital Arts students in Austria, most rejected Midjourney as a main way to make images. They said photography depends on engaging with the real world (EDULEARN26 paper). Schools, meanwhile, are adding AI modules and usage rules. The FT (10 November 2026) reports that art schools are working out how to teach AI imagery without undermining the reasons people study photography. Only the article's summary was reviewed, and no other outlet has corroborated its school-specific details.

What changed and event timeline

  1. AI image wins, then is refused

    Boris Eldagsen declined a Sony World Photography Award for an AI-generated entry, saying competitions were not ready for AI images. Organisers said he had misled them about how much AI was involved ().

  2. World Press Photo bans generative imagery

    After backlash, the contest banned generative fill and fully generated images in all categories, citing accuracy and trustworthiness ().

  3. Classroom method published

    A University of Gothenburg piece describes a 180-minute exercise. Students alter real (non-AI) photos with generative tools, then critique the results using theory ().

  4. Spéos adds an AI module

    The Paris photography school announced ten AI sessions for its studio programmes from September 2026. Topics include retouching, image generation and automating workflows ().

  5. 1 Jul 2026: Student survey presented

    At EDULEARN26, Upper Austria researchers reported that students saw prompting and photography as fundamentally different practices ().

  6. FT reports on schools' dilemma

    The article covers art schools weighing how to teach AI imagery while keeping photography's core purpose ().

Capabilities and access

  • Midjourney: the text-to-image tool in the Austrian study. The version is not stated (IATED).
  • Adobe Creative Cloud: College for Creative Studies (CCS) lets photography students use its built-in AI tools for small edits to their own images without citing them.
  • Spéos: its module does not name specific tools (Spéos).
Read the full section
  • Midjourney: the text-to-image tool in the Austrian study. The version is not stated (IATED).
  • Adobe Creative Cloud: College for Creative Studies (CCS) lets photography students use its built-in AI tools for small edits to their own images without citing them. Any other AI tool needs written instructor approval for 2026/27 (CCS guidelines).
  • Spéos: its module does not name specific tools (Spéos).

Technical analysis for researchers and developers

  • Study design: The Austrian study surveyed students after they finished conventional instruction (camera, lighting, post-production) and then used Midjourney.
  • Reproducible protocol: The Gothenburg exercise has five timed steps that can be reused. Evidence for it comes from student feedback only (Frontiers).
  • Provenance: CCS requires students to keep unedited originals and to document any additional AI use in their coursework.
Read the full section
  • Study design: The Austrian study surveyed students after they finished conventional instruction (camera, lighting, post-production) and then used Midjourney. It is a single cohort with no control group, so it measures perceptions, not skill outcomes (IATED).
  • Reproducible protocol: The Gothenburg exercise has five timed steps that can be reused. Evidence for it comes from student feedback only (Frontiers).
  • Provenance: CCS requires students to keep unedited originals and to document any additional AI use in their coursework. This builds an audit trail into grading (CCS).

Claims and evidence

  • Students value contact with the real world and accept AI for auxiliary tasks: reported for masking and early ideation, based on 39 respondents at one institution (IATED).
  • Generative outputs reflect statistical averages: this is a student observation in a perspective article, not a measured finding (Frontiers).
  • AI training improves employability: this is the school's own rationale, and no outcome data is published (Spéos).
Read the full section
  • Students value contact with the real world and accept AI for auxiliary tasks: reported for masking and early ideation, based on 39 respondents at one institution (IATED).
  • Generative outputs reflect statistical averages: this is a student observation in a perspective article, not a measured finding (Frontiers).
  • AI training improves employability: this is the school's own rationale, and no outcome data is published (Spéos).
  • The FT's school-specific reporting: independent corroboration is unavailable (FT).

Context and prior work

Authorities in the field have been drawing lines on AI since 2023:

Read the full section

Authorities in the field have been drawing lines on AI since 2023:

  • Competitions: World Press Photo excludes generative content (World Press Photo). The Eldagsen case exposed gaps in how competitions vetted entries (Silicon Republic).
  • Schools: they now span three approaches:
  • permissive-with-disclosure (CCS)
  • integration focused on professional skills (Spéos)
  • critical-literacy exercises (Gothenburg)

Limitations, safety and contested findings

  • Thin evidence: the empirical base is small. It amounts to one 39-person survey and one perspective piece describing a teaching method.
  • Policy differences: schools disagree on what counts as acceptable AI use.
  • Disputed facts: in the Eldagsen case, the artist and the organisers gave conflicting accounts of what judges knew (Silicon Republic).
Read the full section
  • Thin evidence: the empirical base is small. It amounts to one 39-person survey and one perspective piece describing a teaching method.
  • Policy differences: schools disagree on what counts as acceptable AI use. CCS exempts embedded editing tools from citation, while World Press Photo bans generative fill entirely (CCS; World Press Photo).
  • Disputed facts: in the Eldagsen case, the artist and the organisers gave conflicting accounts of what judges knew (Silicon Republic).

Business and practitioner implications

  • Graduate skills: expect graduates trained in AI retouching and production workflows (Spéos) who are also wary of AI as a way to author images.
  • Editorial and journalism buyers: they will need provenance controls that match contest-style bans (World Press Photo).
  • Tool vendors: auxiliary features such as masking and ideation fit what students say they accept better than full generation does (IATED).
Read the full section
  • Graduate skills: expect graduates trained in AI retouching and production workflows (Spéos) who are also wary of AI as a way to author images.
  • Editorial and journalism buyers: they will need provenance controls that match contest-style bans (World Press Photo).
  • Tool vendors: auxiliary features such as masking and ideation fit what students say they accept better than full generation does (IATED).
  • Teams: documented process and kept originals, as CCS requires, are a model they can adopt (CCS).
FOLLOW THE EVIDENCE

The source trail.

Sources (7)
A LITTLE LESS NOISE. A LOT MORE CONTEXT.

Stay curious.
Follow the evidence.

Independent perspectives, the original sources, and room for the questions that don't have easy answers.

How we build the brief
Connect with us

Find us where you already read.