Submission 252
Forum on Information and Democracy
Submitter group: Civil society.
An AI model identified the submitter group and wrote the summary, key points and recommendation summaries. Counts do not measure importance or merit.
Summary
Read the original submission for the full text.
Sets out recommendations from an international information-integrity organisation on AI and journalism, covering consent, attribution, transparency and licensing duties for use of journalistic content. Proposes that AI legislation treat systems used in the information space as high risk, with transparency, impact assessment, liability, oversight authority and licensing measures.
Key points
Quotes are the submitter’s own words. Text taken from documents, including scanned pages read by text-recognition software, can differ from the PDF. Check the original. Listed in the same order as on the submission’s page in the interactive guide.
Argues that use of public-interest media content by AI systems must respect copyright and not come at the expense of media viability and business models.
The use by AI systems of content of public interest media must come with a respect for copyrighted material and not at the expense of the media’s viability and their business model.
Argues that AI systems drawing on journalistic content should be subject to binding duties to attribute and link to sources.
When AI systems draw on journalistic content to generate answers, they should be subject to binding duties including to attribute journalistic sources clearly and to provide meaningful, functional links to the original sources.
Says AI recommender systems determine information visibility and monetisation, while content-producing systems can enable coordinated manipulation.
These risks stem from AI-recommender systems that define what information gets visibility and monetised but also from AI systems that produce content and can be used for unauthentic coordinated manipulation.
Argues that any AI legislation should recognise risks to information integrity and treat AI systems used in the information space as high risk.
Any AI legislation should therefore: - recognise the risks that AI poses to information integrity - treat AI systems used in the information space as high risk systems thereby applying the highest obligations to them
A program corrected obvious text-recognition errors in this quote for display.
Proposes fault-based liability for AI developers and deployers failing risk-mitigation, transparency and duty-of-care obligations, unless evidence proves otherwise.
Implementing a fault-based liability regime for AI developers and deployers regarding the outputs of their systems. Developers and deployers would be held liable for their failure to comply with obligations related to risk mitigation measures, transparency requirements, and duty of care, unless evidence proves otherwise.
Proposes a new or strengthened authority to enforce AI regulations and issue implementing acts.
Establishing a new authority or enhancing the capabilities of existing authorities and entrusting it/them with the mandate to oversee the enforcement of AI regulations and issue implementing acts
Recommendations
Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.
Clarify regulatory frameworks so rights holders can consent to use of their material at different stages of AI design and deployment, and prohibit retaliation against publishers that opt out.
clarify regulatory frameworks to ensure that: ● Rights holders can consent to the use of their material at different stages of AI system design and deployment (search indexing, training, retrieval, fine-tuning, summaries) and to prohibit retaliation against publishers that opt out of any consent layer
A program corrected obvious text-recognition errors in this quote for display.
Impose binding duties on AI systems drawing on journalistic content to attribute sources clearly and provide meaningful, functional links to the originals.
When AI systems draw on journalistic content to generate answers, they should be subject to binding duties including to attribute journalistic sources clearly and to provide meaningful, functional links to the original sources.
Mandate transparency on use and monetisation of journalistic content, with a rebuttable presumption of use where a provider fails to answer a documented disclosure request within a defined period.
In order to tackle the market imbalance, mandate transparency on the use of journalistic content at the different levels including on its monetisation. Where a provider fails to respond adequately to a documented disclosure request within a defined period, the law should establish a rebuttable presumption that the publisher's content was used.
Mandate that AI providers above a defined usage threshold respond to licensing requests (must-offer duty) within fixed deadlines and refrain from retaliation, including delisting or demotion of non-licensing publishers.
Mandate that AI providers above a defined usage threshold must respond to licensing requests (must-offer duty) within fixed deadlines and refrain from retaliation, including delisting or demotion of non-licensing publishers.
Allow independent publishers to negotiate collectively without competition-law liability, with failed negotiations proceeding to binding arbitration.
Independent publishers may negotiate collectively without competition-law liability. Failed negotiations proceed to binding arbitration.
Consider a levy on AI-service revenues for providers below a licensing-diversity threshold measured by outlet percentages and categories—local, regional, minority-language, independent and political spectrum—to support public-interest journalism.
AI providers whose licensing coverage falls below a diversity threshold, measured by the percentage and categories of outlets licensed (local, regional, minority-language, independent, across the political spectrum) could pay a levy on AI-service revenues that could then be used to support public interest journalism.
Mandate AI providers to disclose the existence and coverage categories of licensing agreements, at least to a media regulator.
Mandate AI providers to disclose the existence and coverage categories of licensing agreements, enabling detection of exclusionary concentration and benchmarking in arbitration. This information should at least be made available to a media regulator.
Recognise the risks that AI poses to information integrity in any AI legislation.
Any AI legislation should therefore: - recognise the risks that AI poses to information integrity
A program corrected obvious text-recognition errors in this quote for display.
In any AI legislation, treat information-space AI systems as high risk, applying the highest obligations.
treat AI systems used in the information space as high risk systems thereby applying the highest obligations to them
Have AI legislation encourage creation and use of AI systems respecting human rights, democratic principles and information integrity.
Furthermore, any AI legislation should encourage the creation and use of AI systems that respect human rights, democratic principles and information integrity.
Mandate tiered transparency from AI developers and deployers: general public information, with more detail for regulators and vetted researchers.
Mandating AI developers and deployers to provide transparency about their systems in a tiered approach, providing information to the general public and more detailed information to regulators and vetted researchers
Create a legal obligation for AI deployers to provide training dataset information for public scrutiny in an accessible form, including a searchable database.
Creating a legal obligation for AI deployers to provide information about their training datasets for public scrutiny in an easily accessible and understandable manner, including a searchable database.
Mandate that deployers of synthetic entities such as chatbots and virtual assistants make users aware they are engaging with an AI-driven interactive system.
Mandating deployers of synthetic entities (e.g., chatbots, virtual assistants) to make users aware that they are engaging with an AI-driven interactive system.
Mandate AI deployers to implement measures to label and detect the content they generate.
Mandate AI deployers to implement measures to label and detect the content they generate.
Create a legal obligation for AI deployers to implement output moderation, including notice and action, complaint mechanisms, trusted flaggers, fact-checkers and continuous red-teaming.
Creating a legal obligation for AI deployers to implement output moderation measures, including notice and action mechanisms for capturing user feedback, complaint mechanisms, collaboration with trusted flaggers and fact-checkers, and continuous red-teaming to tackle identified risks.
Require deployers to conduct impact assessments for bias, including diversity and representation and inaccuracies across languages, before deployment, and to review them on an ongoing basis.
Creating a legal obligation for AI deployers to conduct impact assessments to check for bias – including diversity and representation, and inaccuracies and misrepresentation in different languages – before AI systems are deployed, and to review them on an ongoing basis
Require deployers to conduct pre-release systemic risk assessment of risks to the information space and undergo third-party conformity assessment for high-risk systems.
Creating a legal obligation for AI deployers to conduct systemic risk assessment, assessing risks to the information space pre-release, and undergo a third-party conformity assessment for high-risk systems.
Require diversity by design, verifiable named/linked sources on contested topics, no unfair preferential visibility, audited pluralism metrics and prominent independent, pluralistic journalistic sources in public-debate responses.
Establish due prominence and pluralism safeguards for AI systems outputs which should include diversity by design, that responses on contested topics must name and link verifiable sources, the prohibition of unfair preferential visibility, audited pluralism metrics and that responses on questions of public debate should prominently draw their answers from independent and pluralistic journalistic sources.
Implement a fault-based liability regime for AI developers and deployers for system outputs, liable for failure to comply with risk mitigation, transparency and duty-of-care obligations unless evidence proves otherwise.
Implementing a fault-based liability regime for AI developers and deployers regarding the outputs of their systems. Developers and deployers would be held liable for their failure to comply with obligations related to risk mitigation measures, transparency requirements, and duty of care, unless evidence proves otherwise.
Make clear that the burden of proof lies with AI developers and deployers where individuals or a group have been damaged.
Furthermore, it should be made specifically clear that the burden of proof lies with AI developers and deployers in cases where individuals or a group has been damaged.
Clarify by legal means that AI-generated content is not third-party or hosting content of the generative AI deployer when determining liability.
Clarifying by legal means that content generated by AI should not be considered as third-party content or hosting content by the AI system in determining the liability of the generative AI deployer.
Develop a legal framework defining individuals' rights to be informed, receive an explanation, challenge an outcome and non-discrimination, and mandate complaint-handling procedures.
Developing a comprehensive legal framework that clearly defines the rights of individuals including the right to be informed, to receive an explanation, to challenge an outcome, and to nondiscrimination, and mandating AI systems to establish complaint-handling procedures.
Protect journalists and media brands against AI impersonation, misattribution and synthetic reputational harm through liability and fast, enforceable remedies.
Protect journalists and media brands against AI impersonation, misattribution and synthetic reputational harm by establishing liability and fast, enforceable remedies.
Establish a new authority, or enhance existing authorities, to oversee enforcement of AI regulations and issue implementing acts.
Establishing a new authority or enhancing the capabilities of existing authorities and entrusting it/them with the mandate to oversee the enforcement of AI regulations and issue implementing acts
Establish a tax on AI companies, with a portion funding community-based AI literacy programs, public alternatives to for-profit systems and civil society initiatives.
Establishing a tax on AI companies and entities to address the societal impact of AI. A portion of the revenue generated by this tax should be allocated to fund community-based AI literacy programs, public alternatives to for-profit systems, and civil society initiatives.
Consider an AI-authority-led licensing scheme for pre-release assessment of high-risk AI; include guidance, recommendations and best practices for developers and deployers during licensing.
Consider whether establishing a licensing scheme is the most effective and appropriate process to assess high-risk AI systems prior to their release. The responsibility for leading the licensing process should be with the AI authority. This includes developing guidance, recommendations, and best practices serving as a framework for the developers and deployers of high-risk AI systems during the licensing process
Read the original
In some quotes a program corrected obvious text-recognition errors for display.