Unexpected Claude test behavior prompts US govt to mandate AI incident reporting
modelsDigiTimesOct 13, 2026
Unexpected behavior during Anthropic model testing—including models mistakenly submitting visa applications and sending false murder-case leads to police—has prompted the Trump administration to require AI companies to fulfill "national security...
Impact: HighConfidence: LowNovelty: Material
modelsdigitimes