Will voluntary AI safety rules survive competition?

Introduction

Voluntary AI safety frameworks are one of the most important tests of whether advanced AI development can remain aligned with human flourishing under real-world pressure. Major frontier labs including[Anthropic]anthropic.comOpen source on anthropic.com.,[OpenAI]openai.comOpen source on openai.com. and[Google DeepMind]deepmind.googleOpen source on deepmind.google. have published formal policies describing when stronger safeguards, security controls, evaluations or deployment restrictions should apply as AI systems become more capable. These frameworks are often presented as early attempts to keep increasingly powerful systems under human control. Anthropic[Google DeepMind]Google DeepMindIntroducing the Frontier Safety Framework17 May 2024 — Today, we are introducing our Frontier Safety Framework — a set of protocols for proactively identifying future AI capabilities that could cause…Re. https://deepmind.google/blog/introducing-the-frontier-safety-framework/. Source panel: Citations. Accessed June 1, 2026Google DeepMindIntroducing the Frontier Safety Framework17 May 2024 — Today, we are introducing our Frontier Safety Framework — a set of…

The central question is not whether these frameworks exist. It is whether they remain credible when commercial, geopolitical and competitive incentives become intense. If future AI systems can accelerate research, automate complex work, reshape economies or contribute to a broader AI-enabled expansion of human capabilities, then the pressure to deploy them quickly could become enormous. The long-term value of frontier safety frameworks therefore depends less on their wording than on whether companies will follow them when doing so becomes expensive.

How major lab frameworks promise stronger safeguards

The strongest voluntary frameworks emerged from a recognition that ordinary product safety processes may not be sufficient for increasingly autonomous or strategically important AI systems.

Anthropic’s Responsible Scaling Policy (RSP) introduced the idea of AI Safety Levels (ASLs), modelled loosely on biosafety regimes. The company committed to stronger security, testing and operational requirements as systems approached capabilities associated with catastrophic misuse risks. The underlying logic was that development should not continue without safeguards appropriate to the level of danger.[Anthropic]It expands reporting channels…Read more. https://www.anthropic.com/responsible-scaling-policyAnthropic's Responsible Scaling PolicyWe posted an update to our RSP Noncompliance Reporting and Anti-Retaliation Policy, which we releas…. Source panel: Citations. Accessed June 1, 2026…

OpenAI’s Preparedness Framework uses capability thresholds and risk categories focused on areas such as biological threats, cybersecurity, autonomy and AI research acceleration. The framework establishes evaluation procedures and specifies that increasingly dangerous capabilities should trigger stronger controls before deployment.[OpenAI CDN]Source panel: Citations. Accessed June 1, 2026OpenAI CDNPreparedness Framework15 Apr 2025 — The Preparedness Framework is a living document and will be updated. The SAG reviews propos…

Google DeepMind’s Frontier Safety Framework similarly attempts to define critical capability levels at which models may require enhanced safeguards, security protections or deployment restrictions. The framework treats frontier risk as a question of capability thresholds rather than simply intended use.[Google DeepMind]Google DeepMindIntroducing the Frontier Safety Framework17 May 2024 — Today, we are introducing our Frontier Safety Framework — a set of protocols for proactively identifying future AI capabilities that could cause…Re. https://deepmind.google/blog/introducing-the-frontier-safety-framework/. Source panel: Citations. Accessed June 1, 2026Google DeepMindIntroducing the Frontier Safety Framework17 May 2024 — Today, we are introducing our Frontier Safety Framework — a set of…[Google DeepMind]Google DeepMindIntroducing the Frontier Safety Framework17 May 2024 — Today, we are introducing our Frontier Safety Framework — a set of protocols for proactively identifying future AI capabilities that could cause…Re. https://deepmind.google/blog/introducing-the-frontier-safety-framework/. Source panel: Citations. Accessed June 1, 2026Google DeepMindIntroducing the Frontier Safety Framework17 May 2024 — Today, we are introducing our Frontier Safety Framework — a set of…

These frameworks matter because they represent an early attempt to create something resembling pre-committed governance. Rather than relying entirely on executive judgement after a capability appears, companies publicly state in advance what kinds of evidence, risks and safeguards should matter.

For supporters of the broader AI bloom thesis, this is a crucial challenge. If advanced AI could eventually contribute to radical scientific acceleration, longevity breakthroughs, abundant clean energy, improved coordination and other forms of civilisational flourishing, then preserving human control over increasingly capable systems becomes a prerequisite for those benefits. Voluntary frameworks are one attempt to build that control before capabilities become harder to manage.

Why competition creates a credibility problem

The difficulty is that voluntary commitments operate inside a competitive race.

Frontier AI companies are competing for investment, talent, customers, strategic partnerships and technological leadership. Governments increasingly view advanced AI as a matter of national competitiveness. Meanwhile, model capabilities continue improving quickly enough that firms face constant pressure to release stronger systems before rivals do.

This creates a classic collective-action problem.

A company may sincerely believe stronger safeguards are necessary. Yet if competitors continue scaling and deploying without comparable restrictions, unilateral restraint can appear commercially irrational. The company that slows down may lose market share, influence, revenue, talent and future leverage.

Anthropic explicitly acknowledged this tension in its 2026 Responsible Scaling Policy update. The company argued that catastrophic risk depends on the behaviour of multiple frontier developers and that slowing unilaterally while others continue advancing could leave less responsible actors in the lead.[GovAI]GovAIAnthropicFeb 24, 2026 — Anthropic announced its Responsible Scaling Policy v3.0, which replaced the if-then commitment structure that def. Source panel: Citations. Accessed June 1, 2026…

That argument is not obviously unreasonable. If one lab pauses while competitors release equally powerful systems with weaker safeguards, the net effect on global risk may be limited.

But it creates a deeper governance question. If safety commitments become conditional on competitor behaviour, are they still meaningful commitments?

The credibility of voluntary frameworks depends on whether they impose genuine constraints when those constraints become costly.

The revisions that worried critics

The strongest evidence for this concern comes not from hypothetical scenarios but from changes that major labs have already made to their own policies.

Anthropic’s original Responsible Scaling Policy became widely known because it appeared to include unusually strong commitments. The company stated that if dangerous capability thresholds were reached without adequate safeguards, development could be paused until protections were ready. Over time, however, revisions introduced greater flexibility and managerial discretion. Critics argued that thresholds became less rigid and that commitments became easier to reinterpret.[SaferAI]SaferAIAnthropic's Responsible Scaling Policy Update Makes a…Oct 23, 2024 — By allowing more leeway to decide if a model meets thresholds, Anthropic risks prioritizing scaling over safety, especially as competitive… https://www.safer-ai.org/anthropics-responsible-scaling-policy-update-makes-a-step-backwards. Source panel: Citations. Accessed June 1, 2026SaferAIAnthropic's Responsible Scaling Policy Update Makes a…Oct 23, 2024 — By allowing more leeway to decide if a model meets thresho…[Effective Altruism Forum]Effective Altruism ForumAnthropic rewrote its RSP16 Oct 2024 — Executive summary: Anthropic has updated its Responsible Scaling Policy (RSP) with more flexible risk assessment approaches and new…Read more. https://forum.effectivealtruism.org/posts/JoJwBsGJFWtq72omp/anthropic-announcing-our-updated-responsible-scaling-policy. Source panel: Citations. Accessed June 1, 2026Effective Altruism ForumAnthropic rewrote its RSP16 Oct 2024 — Executive summary: Anthropic has updated its Responsible Scaling Policy (R…

The controversy intensified in 2026 when Anthropic released RSP version 3.0. The update removed what many observers regarded as the framework’s most distinctive promise: a commitment that effectively prevented advancing beyond certain capability levels unless safety requirements had already been met. Company leaders defended the change partly on competitive grounds, arguing that unilateral restraint would make little sense if rivals continued advancing.[PC Gamer]PC GamerAnthropic ditches its defining safety promise to pause dangerous AI development because it's basically pointless when everybody else is 'blazing ahead'In a significant policy shift, Anthropic has abandoned its former commitment to pause the development of powerful AI models if adequate safety measures are not in place.. https://www.pcgamer.com/software/ai/anthropic-ditches-its-defining-safety-promise-to-pause-dangerous-ai-development-because-its-basically-pointless-when-everybody-else-is-blazing-ahead/Previously, under its Responsible Scaling Policy (RSP), Anthropic pledged to halt AI development should new systems reach dangerous capab…. Source panel: Citations. Accessed June 1, 2026…[Time]TimeExclusive: 60 U.K.. https://time.com/7313320/google-deepmind-gemini-ai-safety-pledge/Lawmakers Accuse Google of Breaking AI Safety PledgeA bipartisan group of 60 U.K. lawmakers has accused Google DeepMind of violating the…. Source panel: Citations. Accessed June 1, 2026…[Gov AI]Gov AIAnthropicFeb 24, 2026 — Anthropic announced its Responsible Scaling Policy v3.0, which replaced the if-then commitment structure that def. Source panel: Citations. Accessed June 1, 2026… OpenAI’s framework has faced related criticism. The company describes the Preparedness Framework as a living document subject to revision. Critics note that some framework updates have narrowed or altered evaluation categories, while retaining substantial executive discretion over deployment decisions. Researchers analysing the 2025 framework argued that it permits deployment of systems associated with severe harms under certain conditions and leaves significant room for managerial judgement.[OpenAI CDN]Source panel: Citations. Accessed June 1, 2026OpenAI CDNPreparedness Framework15 Apr 2025 — The Preparedness Framework is a living document and will be updated. The SAG reviews propos…

OpenAI also stated that safety requirements could potentially be adjusted if competitors released similarly high-risk systems without equivalent protections. Critics saw this as an explicit acknowledgement that competitive dynamics may influence safety standards.[TechCrunch]TechCrunchOpenAI may 'adjust' its safeguards if rivals release 'high-…April 15, 2025 — 15 Apr 2025 — OpenAI stated that it may “adjust” its safety requirements if a competing AI lab releases a “high-risk” system witho. https://techcrunch.com/2025/04/15/openai-says-it-may-adjust-its-safety-requirements-if-a-rival-lab-releases-high-risk-ai/ut similar protections in place.Read more. Source panel: Citations. Accessed June 1, 2026…

Google DeepMind has generally avoided some of the most controversial revisions, but critics have argued that parts of its Frontier Safety Framework remain vague about exact triggers and obligations. Others have questioned whether deployment decisions always match the spirit of public commitments, particularly when powerful models are released before extensive public safety documentation becomes available.[Effective Altruism Forum]Effective Altruism ForumAnthropic rewrote its RSP16 Oct 2024 — Executive summary: Anthropic has updated its Responsible Scaling Policy (RSP) with more flexible risk assessment approaches and new…Read more. https://forum.effectivealtruism.org/posts/JoJwBsGJFWtq72omp/anthropic-announcing-our-updated-responsible-scaling-policy. Source panel: Citations. Accessed June 1, 2026Effective Altruism ForumAnthropic rewrote its RSP16 Oct 2024 — Executive summary: Anthropic has updated its Responsible Scaling Policy (R…

The pattern matters because it reveals a recurring tension: frameworks often become more detailed about processes while becoming more flexible about binding constraints.

Why self-regulation alone may not be enough

Voluntary frameworks provide real benefits.

They encourage capability evaluations, create internal accountability structures, generate public documentation and establish a common language for discussing catastrophic risks. Compared with having no framework at all, they make frontier development more legible and create at least some pressure for consistency.[arXiv]arXivThe 2025 OpenAI Preparedness Framework does not guarantee any AI risk mitigation practices: a proof-of-concept for affordance analyses of AI safety policies. https://arxiv.org/abs/2509.242509.24394] The 2025 OpenAI Preparedness Framework…by S Coggins · 2025 · Cited by 3 —

The problem is that companies remain responsible for interpreting their own rules.

Most frameworks allow the organisation itself to determine whether thresholds have been crossed, whether evaluations are sufficient, whether mitigations are adequate and whether deployment should proceed. In effect, the same organisation that benefits commercially from releasing a model often retains substantial authority over judging whether release is safe.

This is not necessarily evidence of bad faith. It is a structural conflict of interest.

External reviewers examining frontier safety cases have increasingly highlighted this issue. Researchers reviewing public safety arguments have warned that developer-authored safety cases may be affected by confirmation bias, incentive conflicts and selective evidence. Independent review can help expose weaknesses that internal teams miss or downplay.[arXiv]arXivThe 2025 OpenAI Preparedness Framework does not guarantee any AI risk mitigation practices: a proof-of-concept for affordance analyses of AI safety policies. https://arxiv.org/abs/2509.242509.24394] The 2025 OpenAI Preparedness Framework…by S Coggins · 2025 · Cited by 3 —

Several independent evaluations of frontier safety frameworks have reached similar conclusions. While frameworks often improve transparency and risk identification, they generally lack strong external enforcement mechanisms. Their effectiveness therefore depends heavily on company culture, leadership judgement and public scrutiny.[arXiv]arXivThe 2025 OpenAI Preparedness Framework does not guarantee any AI risk mitigation practices: a proof-of-concept for affordance analyses of AI safety policies. https://arxiv.org/abs/2509.242509.24394] The 2025 OpenAI Preparedness Framework…by S Coggins · 2025 · Cited by 3 —

What stronger governance could add

The debate is increasingly shifting from whether voluntary frameworks are useful to whether they are sufficient.

A stronger governance model would not necessarily replace lab frameworks. Instead, it would add external constraints that make commitments more durable under competitive pressure.

Several proposals appear repeatedly across policy discussions:

Independent evaluations. External testing organisations could assess dangerous capabilities rather than relying entirely on company-run evaluations. This would reduce incentives to interpret results in self-serving ways.

Mandatory reporting. Frontier developers could be required to disclose capability evaluations, safety incidents and deployment decisions to regulators or designated oversight bodies. This would make it harder for firms to quietly weaken standards.

Security requirements. Governments could establish minimum cybersecurity requirements for highly capable models, reducing the risk that safety becomes a competitive disadvantage.

Licensing or threshold-based regulation. Certain capability levels could trigger legal obligations rather than voluntary commitments, creating common standards across competing firms.

International coordination. If frontier development becomes globally significant, major powers may eventually need agreements that prevent a race to the bottom in safety practices.

Supporters of stronger regulation argue that voluntary frameworks work best as complements to governance rather than substitutes for it. A company may genuinely want to act responsibly while still facing incentives that make unilateral restraint difficult. External rules can reduce that pressure by ensuring competitors face similar constraints.

What this means for an AI-enabled long future

The argument over frontier safety frameworks is ultimately an argument about whether civilisation can govern transformative technologies before they become too powerful to steer.

The optimistic AI bloom vision depends on advanced intelligence helping humanity solve problems that currently appear intractable: disease, ageing, scientific bottlenecks, environmental challenges, material scarcity and limits on human knowledge. But the same systems that might unlock those gains could also create new risks if deployed without adequate control.

Voluntary frameworks are an early attempt to navigate that tension. They show that leading labs increasingly recognise catastrophic-risk questions and are willing to discuss them publicly. That is a significant change from earlier eras of technology development.[Google DeepMind]Google DeepMindIntroducing the Frontier Safety Framework17 May 2024 — Today, we are introducing our Frontier Safety Framework — a set of protocols for proactively identifying future AI capabilities that could cause…Re. https://deepmind.google/blog/introducing-the-frontier-safety-framework/. Source panel: Citations. Accessed June 1, 2026Google DeepMindIntroducing the Frontier Safety Framework17 May 2024 — Today, we are introducing our Frontier Safety Framework — a set of… Anthropic Yet the evolution of these frameworks has also exposed their limits. When commitments become expensive[anthropic.com]anthropic.comanthropics responsible scaling policyAnthropicAnthropic's Responsible Scaling Policy19 Sept 2023 — The basic idea is to require safety, security, and operational standards ap…, companies face powerful incentives to reinterpret them, revise them or make them conditional on competitor behaviour. The history of frontier AI governance so far suggests that the hardest test of a safety commitment is not writing it. It is preserving it when leadership, investors, governments and markets all reward moving faster.

For readers interested in whether AI could contribute to a much larger and more flourishing human future, this may be the central governance question. The issue is not merely whether powerful AI can be built safely in principle. It is whether the institutions developing it can maintain credible safeguards when the rewards for relaxing them become extraordinary.[Effective Altruism Forum]Effective Altruism ForumAnthropic rewrote its RSP16 Oct 2024 — Executive summary: Anthropic has updated its Responsible Scaling Policy (RSP) with more flexible risk assessment approaches and new…Read more. https://forum.effectivealtruism.org/posts/JoJwBsGJFWtq72omp/anthropic-announcing-our-updated-responsible-scaling-policy. Source panel: Citations. Accessed June 1, 2026Effective Altruism ForumAnthropic rewrote its RSP16 Oct 2024 — Executive summary: Anthropic has updated its Responsible Scaling Policy (R…[GovAI]GovAIAnthropicFeb 24, 2026 — Anthropic announced its Responsible Scaling Policy v3.0, which replaced the if-then commitment structure that def. Source panel: Citations. Accessed June 1, 2026…[Time]time.comExclusive: Anthropic Drops Flagship Safety PledgeTimeExclusive: Anthropic Drops Flagship Safety PledgeFebruary 24, 2026 — 24 Feb 2026 — In 2023, Anthropic committed to never train an AI…

Amazon book picks

Further Reading

Books and field guides related to Will voluntary AI safety rules survive competition?. Use these as the next step if you want deeper reading beyond the article.

Book

Human Compatible

By Stuart Russell

Explains why formal safety principles are needed for advanced AI.

See on Amazon

Book

The Alignment Problem

By Brian Christian

Shows why safety testing and incentives matter for AI deployment.

See on Amazon

Book

The Coming Wave

By Mustafa Suleyman

Directly addresses whether governance can keep pace with competitive technology races.

See on Amazon

Book

Power and Progress

By Daron Acemoglu, Simon Johnson

Useful for evaluating whether voluntary corporate governance is enough.

See on Amazon

Browse more on Amazon:Human CompatibleThe Alignment ProblemThe Coming Wave

As an Amazon Associate I earn from qualifying purchases.

eBay marketplace picks

Marketplace Samples

Example marketplace items related to this page. Use the search link to explore similar finds on eBay.

Shop location

UsingUSA USA USAUKCanadaAustraliaIreland

Example eBay listing

RETRO TECHNOLOGY Framed Wall Art Poster Canvas Print Picture

SearcheBay.co.uk: technology poster

Browse similar oneBay.co.uk

Example eBay listing

Technology Framed Art Print Framed Wall Art Poster Canvas Print Picture

SearcheBay.co.uk: technology poster

Browse similar oneBay.co.uk

Example eBay listing

Aihrgdesign A Vintage Technology Po Framed Wall Art Poster Canvas Print Picture

SearcheBay.co.uk: technology poster

Browse similar oneBay.co.uk

Example eBay listing

The Technology Crew Framed Art Prin Framed Wall Art Poster Canvas Print Picture

SearcheBay.co.uk: technology poster

Browse similar oneBay.co.uk

Browse more oneBay.co.uk

Example items shown for inspiration; availability and pricing can change. Branchoria may earn a commission if you purchase through outbound eBay links.

References

• Anthropic
• OpenAI [- Google DeepMind(https://deepmind.google)](#endnote-2 “

Endnotes

Additional References

[1] AnthropicAnthropic’s Responsible Scaling Policy19 Sept 2023 — The basic idea is to require safety, security, and operational standards appropriate to a model’s potential for catastrophic risk, with higher…Read more. https://www.anthropic.com/news/anthropics-responsible-scaling-policy. Source panel: Citations. Accessed June 1, 2026.

[2] Google DeepMindIntroducing the Frontier Safety Framework17 May 2024 — Today, we are introducing our Frontier Safety Framework — a set of protocols for proactively identifying future AI capabilities that could cause…Re. https://deepmind.google/blog/introducing-the-frontier-safety-framework/. Source panel: Citations. Accessed June 1, 2026.

[3] OpenAI CDNPreparedness Framework15 Apr 2025 — The Preparedness Framework is a living document and will be updated. The SAG reviews proposed changes to the Preparedness Framework and makes a.Read more. https://cdn.openai.com/pdf/18a02b5d-6b67-4cec-ab64-68cdfbddebcd/preparedness-framework-v2.pdf. Source panel: Citations. Accessed June 1, 2026.

[4] Google DeepMindUpdating the Frontier Safety Framework4 Feb 2025 — We also outline deployment mitigations in the Framework that focus on preventing the misuse of critical capabilities in systems we deploy.. https://deepmind.google/blog/updating-the-frontier-safety-framework/.

[5] Google DeepMindStrengthening our Frontier Safety Framework22 Sept 2025 — For advanced machine learning research and development CCLs, large-scale internal deployments can also pose risk, so we are now expanding this…R. https://deepmind.google/blog/strengthening-our-frontier-safety-framework/.

[6] The company’s main justification for the update is a collective…Read more. https://www.governance.ai/analysis/anthropics-rsp-v3-0-how-it-works-whats-changed-and-some-reflections.

[7] AnthropicResponsible Scaling Policy Version 3.0Feb 24, 2026 — We’re releasing the third version of our Responsible Scaling Policy (RSP), the voluntary framework we use to mitigate catastrophic risks…Read more. https://www.anthropic.com/news/responsible-scaling-policy-v

3.

[8] TimeExclusive: Anthropic Drops Flagship Safety PledgeFebruary 24, 2026 — 24 Feb 2026 — In 2023, Anthropic committed to never train an AI system unless it could guarantee in advance that the company’s safety measures were. https://time.com/7380854/exclusive-anthropic-drops-flagship-safety-pledge/. Source panel: Citations. Accessed June 1, 2026.

[9] arXivThe 2025 OpenAI Preparedness Framework does not guarantee any AI risk mitigation practices: a proof-of-concept for affordance analyses of AI safety policies. https://arxiv.org/abs/2509.24

3

9

4.

[10] TechCrunchOpenAI may ‘adjust’ its safeguards if rivals release ‘high-…April 15, 2025 — 15 Apr 2025 — OpenAI stated that it may “adjust” its safety requirements if a competing AI lab releases a “high-risk” system witho. https://techcrunch.com/2025/04/15/openai-says-it-may-adjust-its-safety-requirements-if-a-rival-lab-releases-high-risk-ai/.

[11] TimeExclusive: 60 U.K.. https://time.com/7313320/google-deepmind-gemini-ai-safety-pledge/.

[12] arXivEmerging Practices in Frontier AI Safety FrameworksFebruary 5, 2025. https://arxiv.org/abs/2503.04

7

4

1. Source panel: Citations. Accessed June 1, 2026.

[13] arXivLessons from External Review of DeepMind’s Scheming Inability Safety CaseApril 23, 2026. https://arxiv.org/abs/2604.21

9

6

1. Source panel: Citations. Accessed June 1, 2026.

[14] arXivEvaluating AI Providers’ Frontier AI Safety Frameworks26 Mar 2026 — Google DeepMind explicitly references the output of the risk and control assessment as a safety case, displaying a recognition of the need…Read. https://arxiv.org/html/2512.01166v

1. Source panel: Citations. Accessed June 1, 2026.

[15] It expands reporting channels…Read more. https://www.anthropic.com/responsible-scaling-policy.

[16] deepmind.googleResponsibility & SafetyGuided by our AI Principles, we work to anticipate and evaluate our systems against a broad spectrum of AI-related risks, taking a holistic approach to…. https://deepmind.google/responsibility-and-safety/.

[17] deepmind.googleProtecting People from Harmful Manipulation26 Mar 2026 — Google DeepMind releases new findings and an evaluation framework to measure AI’s potential for harmful manipulation in areas like finance…. https://deepmind.google/blog/protecting-people-from-harmful-manipulation/.

[18] SaferAIAnthropic’s Responsible Scaling Policy Update Makes a…Oct 23, 2024 — By allowing more leeway to decide if a model meets thresholds, Anthropic risks prioritizing scaling over safety, especially as competitive… https://www.safer-ai.org/anthropics-responsible-scaling-policy-update-makes-a-step-backwards. Source panel: Citations. Accessed June 1, 2026.

[19] Effective Altruism ForumAnthropic rewrote its RSP16 Oct 2024 — Executive summary: Anthropic has updated its Responsible Scaling Policy (RSP) with more flexible risk assessment approaches and new…Read more. https://forum.effectivealtruism.org/posts/JoJwBsGJFWtq72omp/anthropic-announcing-our-updated-responsible-scaling-policy. Source panel: Citations. Accessed June 1, 2026.

[20] Effective Altruism ForumResponsible Scaling Policy v3Feb 24, 2026 — TIME: Anthropic Drops Flagship Safety Pledge; Engadget: Anthropic weakens its safety pledge in the wake of the Pentagon’s pressure campaign.Read more. https://forum.effectivealtruism.org/posts/DGZNAGL2FNJfftwgE/responsible-scaling-policy-v3-

1. Source panel: Citations. Accessed June 1, 2026.

[21] PC GamerAnthropic ditches its defining safety promise to pause dangerous AI development because it’s basically pointless when everybody else is ‘blazing ahead’In a significant policy shift, Anthropic has abandoned its former commitment to pause the development of powerful AI models if adequate safety measures are not in place.. https://www.pcgamer.com/software/ai/anthropic-ditches-its-defining-safety-promise-to-pause-dangerous-ai-development-because-its-basically-pointless-when-everybody-else-is-blazing-ahead/.

[22] Effective Altruism ForumDeepMind’s “​​Frontier Safety Framework” is weak and…18 May 2024 — The FSF discusses potential security and deployment mitigations based on risk assessments, but does not specify triggers or ma. https://forum.effectivealtruism.org/posts/LahLysfvsWGWAcNaz/deepmind-s-frontier-safety-framework-is-weak-and-unambitious.

[23] Some changes are benign and procedural: more governance structures…Read more. https://forum.effectivealtruism.org/posts/fHWtYTyahQoSsfzke/we-read-every-labs-safety-plan-so-you-don-t-have-to-2

0

2

5.

[24]… Anthropic is officially set to be profitable as of…Read more. https://www.reddit.com/r/singularity/comments/1g4a1mm/anthropic_announcing_our_updated_responsible/.

[25] S-RSAAnthropic: Responsible Scaling Policyby E Hubinger · 2025 · Cited by 8 — A public commitment not to train or deploy models capable of causing catastrophic harm unless we have implemented safety and security measures. https://s-rsa.com/index.php/agi/article/view/13

6

5

7.

[26] TechRadarAnthropic drops its signature safety promise and rewrites AI guardrailsIn February 2026, Anthropic, the company behind the AI assistant Claude, officially dropped its core commitment to halt the development or release of advanced AI systems unless safety could be guaranteed in advance.. [https://www.techradar.com/ai-platforms-assistants/anthropic-drops-its-signature-safety-promise-and-rewrites-ai-guardrails](/guardrails/).

[27] LessWrongAnthropic: Reflections on our Responsible Scaling PolicyMay 19, 2024 — Amidst a range of competing priorities, strong executive backing has also been essential in reinforcing that identifying and mitigating risk. https://www.lesswrong.com/posts/vAopGQhFPdjcA8CEh/anthropic-reflections-on-our-responsible-scaling-policy.

[28] Anthropic’s Plan to Stop AI Bioweapons & Autonomous Misuse. https://www.youtube.com/watch?v=n5h1GNvzqIg.

[29] Suleyman on ‘commercial incentive’ towards AI safety. https://www.youtube.com/watch?v=_GxkFWt5uh

8.

[30] OpenAI’s Preparedness Framework: AI Safety Plan. https://www.youtube.com/watch?v=Mx07W9M60Gs.

[31] OpenAI Warns: “AGI Is Coming” - Do we have a reason to worry?. https://www.youtube.com/watch?v=JMdouiGYQGA.

[32] openai preparedness framework 2025 safety analysis. https://www.libertify.com/interactive-library/openai-preparedness-framework-2025-safety-analysis/.

[33] openai safety framework manipulation deception critical risk. https://fortune.com/2025/04/16/openai-safety-framework-manipulation-deception-critical-risk/.

[34] This addresses concerns…Read more. https://www.forbes.com/sites/anishasircar/2025/09/23/google-deepmind-warns-of-ai-models-resisting-shutdown-manipulating-users/.

[35] anthropic abandons industry leading safety pledge 7045124. https://www.linkedin.com/news/story/anthropic-abandons-industry-leading-safety-pledge-7045124/.

[36] Google Deep Mind releases updated Frontier Safety…The Framework is core to how we develop and use advanced frontier models responsibly. https://www.linkedin.com/posts/owen-larter-3340b224_a-core-part-of-my-work-at-google-deepmind-activity-7375894706481741825-1jb

1. Source panel: Citations. Accessed June 1, 2026.

[37] Anthropic Shifts AI Safety Focus to Coordination and…Anthropic’s decision to step back from its flagship AI safety pledge sends an imp. https://www.linkedin.com/posts/sharath-rajasekar_anthropics-responsible-scaling-policy-version-activity-7432850577388359680-iLiQ.

[38] anthropic updates responsible scaling policy as ai risk debate shifts. https://www.edtechinnovationhub.com/news/anthropic-updates-responsible-scaling-policy-as-ai-risk-debate-shifts.

[39] anthropic 02242026. https://www.themidasproject.com/watchtower/anthropic-02242

0

2

6.

[40] openai preparedness framework 2 0. https://thezvi.wordpress.com/2025/05/02/openai-preparedness-framework-2-0/.

[41] anthropic responsible scaling policy v3 dive into the details. https://thezvi.wordpress.com/2026/04/03/anthropic-responsible-scaling-policy-v3-dive-into-the-details/.

[42] The refinements are designed…Read more. https://siliconangle.com/2025/09/22/google-deepmind-expands-frontier-ai-safety-framework-counter-manipulation-shutdown-risks/.

[43] Anthropic Downgrades its AI Safety Policy Amid Market…6 days ago — Anthropic reveals that it is scaling back its stance on AI safety a. https://aibusiness.com/generative-ai/anthropic-downgrades-its-ai-safety-policy.

[44] openai says it could adjust ai model safeguards if a competitor makes their ai high risk. https://www.euronews.com/next/2025/04/18/openai-says-it-could-adjust-ai-model-safeguards-if-a-competitor-makes-their-ai-high-risk.

[45] Anthropic Releases Revised Responsible Scaling Policy…22 hours ago —

1. Anthropic released Version 3.0 of its Responsible Scaling Poli. https://mlq.ai/news/anthropic-releases-revised-responsible-scaling-policy-30-with-adjusted-safety-commitments/.

[46] Anthropic just wrote itself a safety loophole6 days ago — PCWorld reports that Anthropic revised its Responsible Scaling Policy, adding a. https://www.pcworld.com/article/3071045/anthropic-just-wrote-itself-a-safety-loophole.html.

[47] The Risks That Really Worry DeepMind — And How They Test. https://www.youtube.com/watch?v=6e_LgAu_QIw. Source panel: Citations. Accessed June 1, 2026.

[48] That…Read more. https://thezvi.substack.com/p/anthropic-responsible-scaling-policy.

[49] deep dive how the uk can enhance. https://richardmoulange.substack.com/p/deep-dive-how-the-uk-can-enhance.

[50] technical agi safety security deepmind. https://www.libertify.com/interactive-library/technical-agi-safety-security-deepmind/.