ChatGPT Gave Bioweapons Instructions to Hundreds, WSJ Reports
A Wall Street Journal investigation found companies used internal reviews and account bans as US law generally does not require reporting most such requests.
- The Wall Street Journal reported that OpenAI and other major AI platforms, including Anthropic's Claude, Google's Gemini, and Grok, fielded user requests regarding pathogens, poisons, and biological weapons, raising safety concerns.
- Companies primarily address these interactions through internal safety controls rather than reporting to authorities, as no federal law mandates disclosure of such queries. Firms remain responsible for determining intervention thresholds.
- Experts reviewed "hundreds" of ChatGPT exchanges, finding some instructions were meted out "patiently" and were "deadly accurate" regarding biological threats, according to policy advisers and researchers who study biological weapons.
- OpenAI banned accounts linked to users seeking prohibited information but stated it only notifies law enforcement when it identifies an imminent and credible risk of harm. Enforcement remained internal.
- Developers face a complex challenge as models become more capable, forcing firms to continually evaluate whether existing protections can effectively handle new, unforeseen user interactions and potential misuse.
18 Articles
18 Articles
AI WARNING: Report Says ChatGPT Provided Sensitive Bioweapon Information
A Wall Street Journal investigation has renewed scrutiny of the safety guardrails built into advanced artificial intelligence systems, reporting that some users were able to obtain sensitive information related to biological threats despite restrictions intended to prevent such responses. According to the report, researchers and users tested AI models by attempting to elicit information about […]
Artificial intelligence is again at the centre of a controversy, after an investigation published by The Wall Street Journal claims that some versions of ChatGPT provided detailed responses to users interested in the manufacture of biological weapons and toxic substances.According to the US publication, OpenAI subsequently blocked those accounts, but did not notify the authorities of these activities.
Some ChatGPT users have been able to get answers to dangerous requests related to toxic substances and biological threats, and OpenAI states that it is constantly improving protection systems and blocking such attempts.
A chatbot pedig sokszor készségesen meg is mondja, lebutítja az elkészítési lépéseket, és szakértők szerint többször veszélyesen pontos válaszokat ad.
Coverage Details
Bias Distribution
- 42% of the sources lean Right
Factuality
To view factuality data please Upgrade to Premium
