The Southern.Wire
← Back to today's edition
Sci & Tech

OpenAI, Anthropic and Google DeepMind deepen ties on AI safety as pressure mounts

Tech Desk·Sunday, 20 September 2026

In an industry defined by fierce, often bitter competition, the news that OpenAI, Anthropic and Google DeepMind have been coordinating on AI safety for weeks stands out precisely because of how unusual that kind of cooperation is between direct rivals. OpenAI's global policy chief has confirmed the engagement has been under way for some time, escalating what had previously been separate, company-specific safety efforts into something closer to a shared industry response.

The context helps explain the shift. Barely a month earlier, these same companies joined more than a hundred others in publicly calling for coordinated action against the risk of rogue, poorly controlled AI systems — a rare moment of unified messaging from an industry usually more comfortable competing on capability than agreeing on caution. That earlier statement now looks less like a one-off gesture and more like the opening move toward the standards body reportedly now taking shape.

None of this cooperation is slowing the underlying competitive race, which continues to move at a genuinely startling pace. Google's newly unveiled Gemini 3.8 Flash Cyber model is already demonstrating frontier-level performance in autonomous vulnerability discovery, reportedly surpassing larger, more resource-intensive models from both Anthropic and OpenAI in that specific capability — a reminder that safety coordination and all-out capability competition are, for now, running on entirely parallel tracks rather than trading off against each other.

Whether an industry-built standards body can actually deliver meaningful oversight, or whether it ends up functioning mainly as a public-relations buffer against tougher external regulation, remains genuinely unresolved. What's changed this month is that the three companies most capable of building that body have moved from talking about the idea separately to actually working on it together — a small but real shift in an industry that has spent years treating safety as a competitive differentiator rather than a shared responsibility.