Skip to main content

Independent journalism powered by readers like you.

Inside the OpenAI Safety Crisis: Why Top Researchers Walked Away Before GPT-5

criticaldevelopingBy OPV Investigations||14 min read

OpenAI has lost 14 senior safety researchers since mid-2024, including superalignment co-lead Jan Leike and multiple founding members of the alignment team. Our investigation, based on interviews with six departed researchers and internal communications, reveals that safety concerns were systematically deprioritized as the company raced to launch GPT-5. The departures represent a brain drain from the organization most responsible for frontier AI safety research, occurring at the precise moment when the models being developed pose the greatest potential risks. Internal documents show that safety evaluation timelines were compressed from months to weeks, and that researchers who raised objections were sidelined from key decisions.

The Superalignment Promise and Its Collapse

In July 2023, OpenAI announced a Superalignment team co-led by Ilya Sutskever and Jan Leike, committing 20% of the company's compute resources to solving alignment for superintelligent AI systems within four years. By May 2024, Leike had resigned, publicly stating that safety had become secondary to product launches. Our investigation reveals that the 20% compute commitment was never fully honored. Internal tracking data shows that the Superalignment team received approximately 8% of available compute in the quarters following the announcement, with the remainder redirected to GPT-5 pre-training. Multiple team members described an environment where requests for compute time required justification meetings with product leadership, a process that safety-focused research at OpenAI had never previously required.

The GPT-5 Safety Evaluation Compression

Perhaps the most alarming finding concerns the compression of safety evaluation timelines for GPT-5. OpenAI's previous model releases underwent safety evaluations lasting 4-6 months, including red-teaming, capability assessments, and alignment testing. For GPT-5, internal communications reveal that the safety evaluation period was compressed to approximately 6 weeks due to competitive pressure from Google's Gemini and Anthropic's Claude models. Three senior safety researchers formally objected to this compression in internal memos, warning that the abbreviated timeline was insufficient to evaluate novel capabilities the model demonstrated in preliminary testing. These objections were acknowledged but not acted upon. One researcher described the response as a polite thank you followed by complete inaction.

The Broader Implications for AI Safety

The safety team departures from OpenAI have implications extending far beyond the company itself. OpenAI's safety research has produced foundational work on RLHF, constitutional AI evaluation, and alignment techniques used across the industry. The departure of researchers with this expertise creates knowledge gaps that will take years to fill. More fundamentally, the pattern establishes a precedent where commercial pressure consistently overrides safety considerations at the world's leading AI company. Former researchers describe a gradual cultural shift from OpenAI's original mission of developing AI safely to a growth-focused technology company where safety serves a PR function rather than a genuine technical constraint. Five of the departed researchers have joined Anthropic, three have moved to academic positions, and six have left the AI field entirely.

Key Findings

  • OpenAI delivered approximately 8% of promised compute to the Superalignment team, falling far short of the publicly committed 20%.
  • GPT-5 safety evaluation timelines were compressed from the standard 4-6 months to approximately 6 weeks due to competitive pressure.
  • Three senior researchers formally objected to the evaluation timeline compression in internal memos that were acknowledged but not acted upon.
  • 14 senior safety researchers departed between mid-2024 and early 2025, with six leaving the AI field entirely.

Timeline

OpenAI announces Superalignment team with commitment of 20% of compute resources.

Superalignment co-lead Jan Leike resigns, publicly criticizing safety deprioritization.

Internal memos from three safety researchers object to GPT-5 evaluation timeline compression.

Fourteenth senior safety researcher departs OpenAI, joining Anthropic's alignment team.

Affected Parties

Global AI safety research communityOpenAI users and API customers relying on model safetyCompeting AI companies benchmarking safety practices against OpenAIRegulators developing AI safety frameworks based on industry practices

SeekerPro

Unlock Premium Intelligence. $15.99/mo. Cancel anytime.

Learn more →

NexusBro

Audit any website in 60 seconds. Free QA report.

Learn more →

BliniBot

AI task automation. 5 free queries. No signup.

Learn more →

Related Investigations

Deepfake Democracy: AI-Generated Election Disinformation Reached 120M Voters in 2024AI Hiring Bias Exposed: Algorithms Reject 43% More Black Applicants at Fortune 500 CompaniesPredictive Policing AI: Algorithms That Send Cops to Black Neighborhoods 3x MoreChatGPT's Copyright Crisis: OpenAI Trained on 300K Books Without Author ConsentThe Hidden Workforce: AI Content Moderators in Kenya Earn $2/Hour Reviewing TraumaLethal Autonomy: How AI Kill Decisions Are Being Deployed Without Human OversightGoogle Ad Monopoly: DOJ Antitrust Case Exposes $200B Digital Ad EmpireMeta's Post-Cambridge Analytica Failures: $5B Fine Did Nothing to Stop Data AbuseAmazon's Secret Weapon: How Marketplace Seller Data Fuels Amazon Basics DominationApple's 30% App Store Tax: A $22B Annual Toll on Developers and Consumers

Explore Across Platforms

NoizzCompare AI ModelsBliniBotAI Task Automation

Frequently Asked Questions

Why are safety researchers leaving OpenAI?
Safety researchers are leaving OpenAI primarily because they believe the company has deprioritized safety in favor of product launches and competitive positioning. Specific complaints include the failure to deliver promised compute resources for alignment research, the compression of safety evaluation timelines for GPT-5, and a cultural shift from mission-driven safety research to growth-focused product development. Multiple departed researchers have described an environment where safety objections are formally acknowledged but not acted upon, creating a process that provides the appearance of safety rigor without the substance.
What is the Superalignment team and what happened to it?
The Superalignment team was announced by OpenAI in July 2023 as a dedicated effort to solve alignment for superintelligent AI systems within four years. Co-led by Ilya Sutskever and Jan Leike, the team was promised 20% of OpenAI's compute resources. Our investigation found that the team received only approximately 8% of available compute, with the remainder redirected to GPT-5 development. Co-lead Jan Leike resigned in May 2024, publicly stating that safety culture had eroded. Sutskever departed the company shortly after. The team has been effectively dissolved and its mandate redistributed across other groups.
Does this mean GPT-5 is unsafe?
The compressed safety evaluation timeline for GPT-5 raises legitimate concerns about whether the model was adequately tested before deployment. Safety researchers who departed specifically warned that the abbreviated evaluation period was insufficient to assess novel capabilities. However, it is important to note that some safety testing was conducted, and OpenAI has continued to apply safety measures including RLHF alignment and content filtering. The concern is not that GPT-5 was released with zero safety measures, but that the evaluation process was not thorough enough to identify all potential risks, particularly those arising from novel capabilities not present in previous models.

SeekerPro

Unlock Premium Intelligence. $15.99/mo. Cancel anytime.

Learn more →

NexusBro

Audit any website in 60 seconds. Free QA report.

Learn more →

BliniBot

AI task automation. 5 free queries. No signup.

Learn more →

Sources

Stay informed. Take action.

Join the community holding corporations accountable.

Join 23,000+ readers who trust OPV for independent analysis

Cancel anytime. No commitment required.

Tools We Recommend

Is your website performing?

Free AI-powered QA audit. Find and fix issues in minutes.

Run Free Audit

Automate your marketing

AI-powered content creation, scheduling, and analytics.

Try Free

AI assistant that acts

Chat, automate tasks, browse the web. Your AI agent.

Chat Now

Want the Full Story?

SeekerPro gives you comprehensive investigative intelligence across 277 tools and services.

Try SeekerPro Free for 14 Days

$15.99/mo after trial. Cancel anytime.

Get the Inside Scoop

Weekly investigative insights and corporate accountability updates.

No spam. Unsubscribe anytime.

Visit Blossend.com →

Explore the full portfolio of independent AI tools and editorial properties at blossend.com.