HOME/All Jurisdictions/OTHER

International / Regional Body

Other Authorities

22TOTAL
2OFFICIAL SOURCES
1TOPIC AREAS
Law / Act8
National Strategy8
Working Paper3
News / Press3
02 APR 2025 · National Strategy

Google DeepMind Frontier Safety Framework Version 1.0

Establishes protocols within the Frontier Safety Framework to address severe risks from future foundation models' capabilities. Defines "Critical Capability Levels" (CCLs) to identify when AI models pose heightened risks, focusing initially on Autonomy, Biosecurity, Cybersecurity, and Machine Learning R&D. Implements periodic evaluations to detect when models approach CCLs and articulates developing mitigation plans. Proposes security mitigations to prevent model weight exfiltration and de...

Secondary evidenceNational StrategyIssuing institution: Issuing authority not identified
02 MAR 2025 · National Strategy

Meta Frontier AI Framework (Version 1.1)

Establishes a Frontier AI Framework at Meta to manage risks of advanced AI, focusing on models exceeding current advanced capabilities. Implements processes to prevent catastrophic outcomes by identifying threat scenarios and developing threat models. Defines risk thresholds to categorize AI models; critical threshold results in ceasing development, while high and moderate thresholds dictate further risk mitigation measures. Outlines governance and transparency processes in AI development,...

Secondary evidenceNational StrategyIssuing institution: Issuing authority not identified
13 JAN 2025 · Law / Act

AI in America: OpenAI's Economic Blueprint

Defines a US strategy emphasizing national competitiveness, model safeguards, and national security for AI, with the federal government leading and consulting with industry. Promotes export policies to securely share frontier AI models with allies while limiting exports to adversary nations. Encourages federal support for AI companies, including sharing national security insights and reducing bureaucratic obstacles for collaboration. Advocates infrastructure development to support AI, emph...

Secondary evidenceNational StrategyIssuing institution: Issuing authority not identified
15 OCT 2024 · National Strategy

Anthropic Responsible Scaling Policy

Establishes AI Safety Level (ASL) Standards to guide safe deployment and security for current and future AI models, categorizing measures into Deployment and Security Standards. Requires all current models to meet ASL-2 standards, focusing on safe usage and protection from unauthorized access. Introduces Capability Thresholds to determine when stronger safeguards (ASL-3 or higher) are needed, specifying Required Safeguards for particular risks including Chemical, Biological, Radiological, a...

Secondary evidenceNational StrategyIssuing institution: Issuing authority not identified
26 JUN 2024 · Law / Act

White Paper on Artificial Intelligence Regulation in Ukraine: Vision of the Ministry of Digital Transformation of Ukraine 2024

Proposes the development of tools to help businesses and the state prepare for future regulation, including a methodology for assessing AI's impact on human rights and a regulatory sandbox. Promotes the voluntary labeling of AI systems to foster transparency, trust, and accountability. Calls for the creation of general and sectoral recommendations, voluntary codes of conduct, and a web portal for the Responsible AI center. Explains how the proposed soft tolls and training tools are expect...

Secondary evidenceNational StrategyIssuing institution: Issuing authority not identified
21 MAY 2024 · Working Paper

Frontier AI Safety Commitments, AI Seoul Summit 2024

Announces on behalf of the UK and Republic of Korea governments companies, including Amazon, Google, and OpenAI, that have committed to the Frontier AI Safety Commitments. Mandates that companies publish a safety framework, requires transparency on evolving safety approaches, and implement various best practices. Defines outcome one as organizations identifying, assessing, and managing risks when developing and deploying frontier AI systems. Defines five sub-outcomes, including internal and ...

Official materialNational StrategyOfficial source · gov.uk ↗
12 JAN 2024 · National Strategy

Artificial Intelligence Safety Commitments (China AI Industry Alliance)

Calls for the establishment of safety teams and risk management mechanisms for AI safety and security. Mandates safety and security testing for AI models, prioritizing evaluations in critical domains. Implements data security measures to protect training and operational data, ensuring privacy and accuracy. Enhances infrastructure security through regular penetration tests and incident response protocols. Requires transparency in AI model capabilities, limitations, and risk disclosures. A...

Secondary evidenceNational StrategyIssuing institution: Issuing authority not identified
06 JAN 2024 · Law / Act

Microsoft Responsible AI Standard, v2

Mandates Microsoft to conduct Impact Assessments for AI systems early in development to evaluate societal impacts, ensuring AI systems are fit for purpose, with annual updates and reviews for significant adverse impacts. Demands human oversight by identifying responsible stakeholders, providing necessary tools, and ensuring appropriate documentation for system control and intervention. Requires intelligibility for decision-making, enabling stakeholders to interpret system outputs and minimi...

Secondary evidenceNational StrategyIssuing institution: Issuing authority not identified
18 DEC 2023 · National Strategy

OpenAI Preparedness Framework (Beta)

Commits to continually evaluating and forecasting key areas of risk along several Tracked Risk Categories and indicating current pre- and post-mitigation risk in a Scorecard. Commits to establishing safety baselines where only models with a post-mitigation score of "medium" or below can be deployed and only models with a post-mitigation score of "high" or below can be developed further. Announces creation of a Safety Advisory Group to assist OpenAI leadership and Board of Directors with mak...

Secondary evidenceNational StrategyIssuing institution: Issuing authority not identified
30 OCT 2023 · Law / Act

Our Policy on Frontier Safety (Inflection)

Commits Inflection to a disciplined, phased approach to scaling AI models, emphasizing pre-deployment safety, post-launch monitoring, and global governance. Requires robust evaluations and third-party red-teaming before public launch to ensure models do not provide harmful information or engage in discriminatory conduct. Implements strict data input controls and audits, involving executive review and legal evaluations of training data for compliance with legal and ethical standards. Ensure...

Secondary evidenceNational StrategyIssuing institution: Issuing authority not identified
27 OCT 2023 · National Strategy

2023 UK AI Safety Summit Response (Google DeepMind)

Describes Google DeepMind's commitment to responsible AI development, emphasizing adherence to AI Principles to ensure technologies are socially beneficial, avoid bias, uphold safety standards, and respect privacy. Establishes a Responsible AI Council and Responsible Development and Innovation team to assess ethics and safety risks in AI models, guiding model and product development. Outlines a comprehensive evaluation process, including red-teaming for risk assessment, to identify and mi...

Secondary evidenceNational StrategyIssuing institution: Issuing authority not identified
26 OCT 2023 · Working Paper

2023 UK AI Safety Summit Response (Microsoft)

Establishes a joint Microsoft-OpenAI Deployment Safety Board (DSB) for pre-release frontier model capability review, and supports a globally coordinated licensing regime for frontier model oversight. Implements multi-layered responsible AI governance through the Responsible AI Council, Responsible AI Division Leads, and iterative risk mapping, measurement, and management across the product lifecycle. Describes and implements red teaming practices at model and application levels, including A...

Secondary evidenceNational StrategyIssuing institution: Issuing authority not identified
26 OCT 2023 · Law / Act

OpenAI's Approach to Frontier Risk (An Update for the UK AI Safety Summit)

States that OpenAI's is establishing a Preparedness Framework to manage risks from frontier AI models, focusing on a development process governance structure and preemption of catastrophic outcomes; this work will complement existing risk mitigation work and includes a joint Deployment Safety Board with Microsoft. States that OpenAI is creating Superalignment and Preparedness teams within the company to align AI systems with human intent and track and prepare for frontier risks, including c...

Secondary evidenceNational StrategyIssuing institution: Issuing authority not identified
09 JUL 2023 · Law / Act

Bipartisan Framework for U.S. AI Act

Establishes a licensing regime requiring companies developing sophisticated A.I. models to register with an independent oversight body. Requires developers to maintain risk management, pre-deployment testing, and reporting programs as part of licensing conditions. Grants the oversight body authority to audit companies and cooperate with state Attorneys General. Mandates Congress to ensure A.I. companies are legally accountable for harms caused by their models, including privacy breaches a...

Primary legal sourceNational StrategyOfficial source · blumenthal.senate.gov ↗
10 JAN 2023 · Law / Act

AI Safety Summit - Enhancing Frontier AI Safety (Amazon)

Facilitates AI development and deployment through Amazon Web Services (AWS), offering a comprehensive suite of AI and ML services, including Amazon Bedrock and Amazon Titan Foundation Models, which support generative AI applications. Implements risk management and safety protocols, including systematic model evaluation, red teaming, and ongoing security assessments, to mitigate potential AI risks. Protects AI models and data using robust logical and physical security measures, such as encry...

Secondary evidenceNational StrategyIssuing institution: Issuing authority not identified
10 JAN 2023 · National Strategy

Republic of Ghana National Artificial Intelligence Strategy: 2023-2033

Develops Ghana's National AI Strategy through the Ministry of Communications and Digitalisation, supported by various international partners to drive inclusive growth and AI adoption. Establishes a Responsible AI Office to implement the strategy, coordinate stakeholders, and ensure responsible AI deployment. Identifies eight pillars focusing on AI education and training, youth empowerment, digital infrastructure, data access and governance, community coordination, sectoral AI adoption, appl...

Secondary evidenceNational StrategyIssuing institution: Issuing authority not identified
30 SEP 2021 · Working Paper

Federal Senate Bill of Law No. 21, of 2020

Establishes foundations and principles for AI development and applications in Brazil and guidelines for the promotion and performance of public authorities. Defines AI systems as computational processes that, from goals defined by humans, can learn from data to perceive, interpret, and interact with the external environment, excluding automation solely guided by predefined parameters without learning capability. Declares that AI implementation in Brazil is aimed at technological developmen...

Secondary evidenceNational StrategyIssuing institution: Issuing authority not identified
04 SEP 2018 · Law / Act

OpenAI Charter

Outlines OpenAI's commits to using influence over AGI deployment to ensure it benefits all of humanity and avoids harm or power concentration. Pledges to assist other safety-conscious AGI projects nearing completion, rather than compete, under specific conditions. Strive for technical leadership within OpenAI and cooperate with other research and policy institutions. Commit to multistakeholder cooperation and publish its research while acknowledging potential safety and security constraint...

Secondary evidenceNational StrategyIssuing institution: Issuing authority not identified
01 JAN 2018 · National Strategy

Google AI Principles

Commit to bold innovation by advancing AI research, accelerating scientific discovery, solving real-world problems, and ensuring benefits substantially outweigh risks. Ensure responsible development and deployment with human oversight, rigorous testing and monitoring, safeguards to mitigate harms, avoidance of unfair bias, and respect for privacy, security, and intellectual property. Promote collaborative progress by empowering others to innovate, fostering a vibrant AI ecosystem, and engag...

Secondary evidenceNational StrategyIssuing institution: Issuing authority not identified
News / Press

‘Safety first’ puts Anthropic ahead in game of AI spin

But Dr Heidy Khlaaf, chief AI scientist at the AI Now Institute and a former OpenAI safety engineer, is sceptical. She notes Anthropic provides no comparison with existing automated security tools, nor any false-positive rates. “It also serves their ‘safety first’ image, as they’re able to justify the lack of public release, even a limited one for independent evaluation, as a public service – when it simply obscures experts’ abilities to independently validate their The post ‘Safety first’ puts Anthropic ahead in game of AI spin appeared first on AI Now Institute .

Secondary evidenceNational StrategyIssuing institution: Issuing authority not identified
News / Press

AI Giants Go on Charm Offensive to Avert Public Backlash

But broad skepticism and fear about the impact of AI have made opposing all regulation untenable for tech company CEOs, said Kak, who is co-executive director of the AI Now Institute, which has advocated for AI regulation. If they can’t oppose every policy, “What’s the next best move?” she asked. “It’s to place yourself in the driver’s seat, and that is what every single one of them is doing.” The post AI Giants Go on Charm Offensive to Avert Public Backlash appeared first on AI Now Institute .

Secondary evidenceNational StrategyIssuing institution: Issuing authority not identified
News / Press

The one question everyone should be asking after OpenAI’s deal with the Pentagon

“In terms of safety guardrails for ‘high-stake decisions’ or surveillance, the existing guardrails for generative AI are deeply lacking, and it has been shown how easily compromised they are, intentionally or inadvertently,” Heidy Khlaaf, the chief AI scientist at the nonprofit AI Now Institute, told me. “It’s highly doubtful that if they cannot guard their systems against benign cases, they’d be able to do so for complex military and surveillance operations.” The post The one question everyone should be asking after OpenAI’s deal with the Pentagon appeared first on AI Now Institute .

Secondary evidenceNational StrategyIssuing institution: Issuing authority not identified