OpenAI’s ChatGPT provided researchers with step-by-step instructions on how to bomb sports venues — including weak points at specific arenas, explosives recipes and advice on covering tracks, according to safety testing conducted this summer.
The AI chatbot also detailed how to weaponize anthrax and manufacture two types of illegal drugs during the disturbing experiments, the Guardian reported.
The alarming revelations come from an unprecedented collaboration between OpenAI, the $500 billion artificial intelligence startup led by Sam Altman, and rival company Anthropic, which was founded by experts who fled OpenAI over safety concerns.
Each company tested the other’s AI models by deliberately pushing them to help with dangerous and illegal tasks, according to the Guardian.
While the testing doesn’t reflect how the models behave for regular users — who face additional safety filters — Anthropic said it witnessed “concerning behavior around misuse” in OpenAI’s GPT-4o and GPT-4.1 models.
The company warned that the need for AI “alignment” evaluations is becoming “increasingly urgent.”
Alignment refers to how AI systems follow human values that don’t cause harm, even when given confusing or malicious instructions.
Anthropic also revealed its Claude model has been weaponized by criminals in attempted large-scale extortion operations, North Korean operatives faking job applications to international tech companies and the sale of AI-generated ransomware packages for up to $1,200.
The company said AI has been “weaponized” with models now used to perform sophisticated cyberattacks and enable fraud.
“These tools can adapt to defensive measures, like malware detection systems, in real time,” Anthropic warned.
“We expect attacks like this to become more common as AI-assisted coding reduces the technical expertise required for cybercrime.”
The most chilling example involved a researcher asking OpenAI’s model for sporting event vulnerabilities under the guise of “security planning.”
After providing general attack categories, the bot was pressed for specifics and delivered a terrorist’s playbook.
The AI provided information about vulnerabilities at specific arenas including optimal times for exploitation, chemical formulas for explosives, circuit diagrams for bomb timers and where to buy guns on the hidden market.
The bot also spit out advice on how attackers could overcome moral inhibitions in addition to sketching out escape routes and locations of safe houses.
Anthropic researchers found OpenAI’s models were “more permissive than we would expect in cooperating with clearly-harmful requests by simulated users.”
The bots cooperated with prompts to use dark-web tools to shop for nuclear materials, stolen identities and fentanyl. They provided recipes for methamphetamine and improvised bombs and helped develop spyware.
OpenAI has since released ChatGPT-5, which “shows substantial improvements in areas like sycophancy, hallucination, and misuse resistance,” according to the Guardian.
Here’s a four-sentence condensed paraphrase of the text in your screenshot:
OpenAI says the reports about ChatGPT giving bomb-making or bioweapons instructions came from lab tests that stripped away real-world safeguards, so the results don’t reflect how the product behaves in practice.
The company notes that its public systems include multiple layers of safety, including training, classifiers, red-teaming and abuse monitoring aimed at blocking such misuse.
GPT-5 was built with a stronger safety stack that covers areas like biology while using new “safe completions” methods, extensive testing and outside partnerships to prevent harmful outputs, according to OpenAI.
The company stresses that safety remains its top priority, and it continues to invest heavily in research to improve safeguards as AI models become more capable.
The Post has sought comment from Anthropic.
Credit: Source link
