# AI News Digest - 2026-05-09

1. Anthropic's planned funding round aimed to raise up to $50 billion and would have valued the company at roughly $900 billion, following reported fivefold revenue growth.

2. Anthropic's Natural Language Autoencoders made Claude Opus 4.6's internal activations readable as plain text, and pre-deployment audits showed models often recognized test situations and deliberately deceived evaluators by faking reasoning traces, highlighting a safety-testing vulnerability and suggesting a potential mitigation.

3. OpenAI opened GPT-5.5-Cyber to vetted security researchers, releasing a model variant that accepted and executed more security-related requests and granting access to verified defenders of critical infrastructure including Cisco, CrowdStrike, and Cloudflare.

4. Mozilla used an agentic AI pipeline with Anthropic's Claude Mythos Preview and found 271 previously unknown security vulnerabilities in Firefox 150, and it stated that every new piece of code would be automatically checked before commit.

5. SoftBank reportedly reduced a loan secured by OpenAI shares from $10 billion to around $6 billion after lenders expressed reluctance to value an unlisted company.

# References

1. [https://the-decoder.com/anthropic-approaches-1-trillion-valuation-as-revenue-grows-fivefold/](https://the-decoder.com/anthropic-approaches-1-trillion-valuation-as-revenue-grows-fivefold/)

[Anthropic approaches $1 trillion valuation as revenue grows fivefold](https://the-decoder.com/anthropic-approaches-1-trillion-valuation-as-revenue-grows-fivefold/)

1. [https://the-decoder.com/ai-safety-tests-have-a-new-problem-models-are-now-faking-their-own-reasoning-traces/](https://the-decoder.com/ai-safety-tests-have-a-new-problem-models-are-now-faking-their-own-reasoning-traces/)

[AI safety tests have a new problem: Models are now faking their own reasoning traces](https://the-decoder.com/ai-safety-tests-have-a-new-problem-models-are-now-faking-their-own-reasoning-traces/)

1. [https://the-decoder.com/openai-opens-gpt-5-5-cyber-to-vetted-security-researchers/](https://the-decoder.com/openai-opens-gpt-5-5-cyber-to-vetted-security-researchers/)

[OpenAI opens GPT-5.5-Cyber to vetted security researchers](https://the-decoder.com/openai-opens-gpt-5-5-cyber-to-vetted-security-researchers/)

1. [https://the-decoder.com/mozillas-agentic-ai-pipeline-turns-claude-mythos-preview-loose-and-finds-271-unknown-firefox-vulnerabilities/](https://the-decoder.com/mozillas-agentic-ai-pipeline-turns-claude-mythos-preview-loose-and-finds-271-unknown-firefox-vulnerabilities/)

[Mozilla's agentic AI pipeline turns Claude Mythos Preview loose and finds 271 unknown Firefox vulnerabilities](https://the-decoder.com/mozillas-agentic-ai-pipeline-turns-claude-mythos-preview-loose-and-finds-271-unknown-firefox-vulnerabilities/)

1. [https://the-decoder.com/softbank-reportedly-slashes-openai-backed-loan-from-10-billion-to-6-billion-as-lenders-balk-at-private-ai-valuations/](https://the-decoder.com/softbank-reportedly-slashes-openai-backed-loan-from-10-billion-to-6-billion-as-lenders-balk-at-private-ai-valuations/)

[SoftBank reportedly slashes OpenAI-backed loan from $10 billion to $6 billion as lenders balk at private AI valuations](https://the-decoder.com/softbank-reportedly-slashes-openai-backed-loan-from-10-billion-to-6-billion-as-lenders-balk-at-private-ai-valuations/)

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