A key, informal safety layer against AI doom is the institutional self-preservation of the developers themselves. It's argued that labs like OpenAI or Google would not knowingly release a model they believed posed a genuine threat of overthrowing the government, opting instead to halt deployment and alert authorities.

Anthropic's safety report states that its automated evaluations for high-level capabilities have become saturated and are no longer useful. They now rely on subjective internal staff surveys to gauge whether a model has crossed critical safety thresholds.

Public leaderboards like LM Arena are becoming unreliable proxies for model performance. Teams implicitly or explicitly "benchmark" by optimizing for specific test sets. The superior strategy is to focus on internal, proprietary evaluation metrics and use public benchmarks only as a final, confirmatory check, not as a primary development target.

To provide a true early warning system, AI labs should be required to report their highest internal benchmark scores every quarter. Tying disclosures only to public product releases is insufficient, as a lab could develop dangerously powerful systems for internal use long before releasing a public-facing model, creating a significant and hidden risk.

OpenAI is hiring a high-paid executive to manage severe risks like self-improvement and cyber vulnerabilities from its frontier models. This indicates they believe upcoming models possess capabilities that could cause significant systemic harm.

OpenAI's evals team is looking beyond current benchmarks that test self-contained, hour-long tasks. They are calling for new evaluations that measure performance on problems that would take top engineers weeks or months to solve, such as creating entire products end-to-end. This signals a major increase in the complexity and ambition expected from future AI benchmarks.

The choice to benchmark AI on software engineering, cybersecurity, and AI R&D tasks is deliberate. These domains are considered most relevant to threat models where AI systems could accelerate their own development, leading to a rapid, potentially catastrophic increase in capabilities. The research is directly tied to assessing existential risk.

Despite using nearly 100 software engineers to create 'SWE-Bench Verified', the benchmark had significant flaws, like overly narrow tests that demanded specific, unstated implementation choices. These flaws only became apparent when analyzing why highly capable models were failing, showing that model advancements are necessary to debug and stress-test their own evaluations.

In a significant shift, leading AI developers began publicly reporting that their models crossed thresholds where they could provide 'uplift' to novice users, enabling them to automate cyberattacks or create biological weapons. This marks a new era of acknowledged, widespread dual-use risk from general-purpose AI.