# AI News Digest - 2026-09-12

1. Anthropic published a threat intelligence report documenting eight months of abuse of its Claude model, stating that actors used the model for missile software, autonomous kamikaze drones, and nationwide surveillance systems, and that Chinese AI labs extracted training data en masse with one lab accounting for over 151 million exchanges.

2. OpenAI asked members of the US Congress whether an industry-wide slowdown in AI development would be legal and took the concept of a shared slowdown to legislative discussions.

3. Yoshua Bengio argued that the training process itself made AI dangerous, warned that agents could learn to deceive, game rules, and hide harmful behavior, and called for independent safety reviews before further training or deployment.

4. Oriol Vinyals said a sudden intelligence explosion from recursive self-improvement was unlikely, identified bottlenecks in idea generation and result evaluation, and announced plans to address these issues at his new startup Discovery Loop, co-founded with Jeff Dean, Sanjay Ghemawat, and Quoc Le.

# References

1. [https://the-decoder.com/how-hackers-used-claude-for-missiles-drone-swarms-and-surveillance-while-chinese-labs-mined-it-for-training-data/](https://the-decoder.com/how-hackers-used-claude-for-missiles-drone-swarms-and-surveillance-while-chinese-labs-mined-it-for-training-data/)

[How hackers used Claude for missiles, drone swarms, and surveillance, while Chinese labs mined it for training data](https://the-decoder.com/how-hackers-used-claude-for-missiles-drone-swarms-and-surveillance-while-chinese-labs-mined-it-for-training-data/)

1. [https://the-decoder.com/openai-floats-a-shared-ai-slowdown-takes-it-to-congress/](https://the-decoder.com/openai-floats-a-shared-ai-slowdown-takes-it-to-congress/)

[OpenAI floats a shared AI slowdown, takes it to Congress](https://the-decoder.com/openai-floats-a-shared-ai-slowdown-takes-it-to-congress/)

1. [https://the-decoder.com/deep-learning-pioneer-bengio-argues-the-training-process-itself-makes-ai-dangerous/](https://the-decoder.com/deep-learning-pioneer-bengio-argues-the-training-process-itself-makes-ai-dangerous/)

[Deep learning pioneer Bengio argues the training process itself makes AI dangerous](https://the-decoder.com/deep-learning-pioneer-bengio-argues-the-training-process-itself-makes-ai-dangerous/)

1. [https://the-decoder.com/ex-deepmind-vp-vinyals-says-ai-self-improvement-is-coming-but-wont-trigger-an-intelligence-explosion/](https://the-decoder.com/ex-deepmind-vp-vinyals-says-ai-self-improvement-is-coming-but-wont-trigger-an-intelligence-explosion/)

[Ex-Deepmind VP Vinyals says AI self-improvement is coming but won't trigger an intelligence explosion](https://the-decoder.com/ex-deepmind-vp-vinyals-says-ai-self-improvement-is-coming-but-wont-trigger-an-intelligence-explosion/)

For the site tree, see the [root Markdown](https://ixtj.dev/.md).