AI Safety

Who Really Controls AI Testing? New Study Reveals Hidden Power Imbalance

Who Really Controls AI Testing? New Study Reveals Hidden Power Imbalance

⚡When AI is tested, who holds the power? The answer affects your safety and trust.

Deep Dive

When you hear that an AI system has been tested and deemed safe, you might assume the testers had real power to set the rules. But a new study shows that's often not true. Many organizations and countries participate in AI evaluations, yet they have little control over how those evaluations are designed, carried out, or paid for. The researchers created a framework called PAAI (Power, Access, Adequacy, Independence) to analyze who really holds the authority in AI testing.

They looked at 12 African AI evaluation projects from 2022 to 2025 and the joint testing exercises of an international network of AI Safety Institutes. They found that African researchers often contribute valuable data, language expertise, and task design, but external institutions keep control over the infrastructure and evaluation frameworks. This means that even when African experts help design tests, they don't have the power to run them or decide what gets published. The study calls this "design authority without infrastructure authority," and it never led to full independence.

The joint testing exercises showed a similar pattern. The institutes that led each exercise also ran the models, while Kenya, the only African member of the network, moved from simple annotation to task design but never got to execute models. This mirrors a broader divide: high-income countries dominate the actual testing, while others are limited to supportive roles. The researchers argue that participating in an evaluation doesn't automatically give you a say in how it's done.

For AI safety to be credible, the study says, access must come with control over priorities, methods, publication, and continuity. In other words, it's not enough to be invited to the table—you need a real voice in deciding what's on the menu. This matters because AI systems are increasingly used in decisions that affect our lives, from healthcare to hiring. If the testing is controlled by a narrow group, the results may not reflect the needs and concerns of everyone. The study's findings highlight the need for fairer and more inclusive AI governance.

Key Points
  • Many countries help test AI but have little power over how tests are designed or run.
  • African researchers often provide data and language skills, but outsiders control the tools and rules.
  • For AI safety to be trustworthy, all participants need real control over testing decisions.

Why It Matters

AI safety tests may not be truly independent, affecting trust in AI systems you use daily.

🔒 FREE VIP ACCESS

Read the full breakdown & get daily AI executive intelligence

Join like-minded builders, researchers, and AI professionals who read The Neural Feed daily. 100% free. No spam.

🔔 BREAKING RELEASE RADAR INSTANT ALERTS • 1-CLICK PAUSE

Get instant breaking AI release alerts by topic

Choose the AI labs, models, or chip tracks you follow. Receive fast alerts only when something major ships — plus our free daily morning briefing.

Includes full article unlock • Only major releases • Unsubscribe anytime
🎁 FREE EXECUTIVE DOWNLOAD ISSUE #21

Download This Week's 8-Page Executive AI Briefing (PDF)

Our curated executive intelligence packet: in-depth breakdowns of frontier models, market movements, plus full-color AI comics.

  • ✓ 8-page formatted executive PDF briefing
  • ✓ Frontier AI models & technical breakthroughs
  • ✓ Full-color satirical AI comic strips
Instant 1-click download • 100% free • Unsubscribe anytime
📬 Get the top 10 AI stories daily