BackAnthropic reports disrupting attempts to use AI for biological weapons research

Developing

Al Jazeera13 minutes agoTech1 min read

Anthropic reports disrupting attempts to use AI for biological weapons research

US tech company says newer Claude models can complete complex scientific work, raising safety concerns.

Quick Look

Anthropic reported disrupting multiple attempts to use its Claude AI models for biological weapons research, missile projects, and espionage, as safety concerns grow over advanced artificial intelligence capabilities.

AI-generated summary

Why It Matters

Anthropic issued a report on Thursday outlining the misuse of its AI models, including Claude, over the past eight months.

Font size

Anthropic has disrupted several attempts to use its artificial intelligence models for research that could help develop biological weapons, the United States technology company has reported.

The firm listed the cases in a report issued on Thursday outlining misuse of its models, including the chatbot Claude, over the past eight months. Concern continues to rise over the potential dangers posed by AI development.

Anthropic said it could not determine whether the research was legitimate or malicious, as inquiry that leads to vaccines can also help engineer dangerous pathogens. The company said it erred on the side of caution.

In one case in May, a scientist sought help with a grant application to engineer mutations to chikungunya – a mosquito-borne virus similar to malaria – to make it more harmful in live animals.

Anthropic said the case was worrying in part because it could tell the work was intended for a military research institute. It banned the accounts involved.

“What we don’t know is if the research was meant to be weaponised,” Jacob Klein, Anthropic’s head of threat intelligence, told The New York Times.

Andrew Weber of the Council on Strategic Risks, who reviewed the report before its release, told the newspaper that access to such models should be limited to trusted researchers.

Anthropic said evaluations of its older models last year found they could not meaningfully assist dangerous biological research, but newer models can complete complex scientific work.

The same report set out bids to use Claude for missile projects and espionage. Two days before the report, a researcher quit Anthropic, saying the race to develop AI could wipe out humanity.

President Donald Trump late on Thursday dismissed the rising concerns regarding the technology.

Speaking to reporters, Trump said he is concerned that the US will be in a “very bad position” if it does not lead the AI race.

Asked if he has concerns that AI could lead to human extinction, he said: “No, I don’t have any.”

Open Questions

Which specific military research institutes attempted to access the AI models?
What additional safeguards will Anthropic implement for newer models?

California Governor Gavin Newsom signed 13 new laws to protect minors online, restricting addictive social media features like infinite scrolling and algorithmic recommendations for users under 16, banning AI-generated sexual content involving minors, requiring parental controls for AI chatbots, and prohibiting AI-powered toys, citing California's approach as more effective than Australia's under-16 social media ban.

Deutsche Welle

2 min read

Developing·3 hours ago

Anthropic Claims It Thwarted Multiple Malicious AI Operations Using Claude Models

Anthropic reported blocking several malicious uses of its Claude AI models, including attempts to develop missile guidance in Yemen, Russian and Chinese cyber-espionage campaigns, Iranian influence operations, and a recruitment scheme targeting Uyghurs in Syria, while also addressing internal safety concerns and legal disputes with the Pentagon.

Al Jazeera

2 min read

Developing·8 hours ago

OpenAI Seeks Congressional Guidance on AI Development Slowdown Amid Antitrust Concerns

OpenAI has asked members of Congress for guidance on whether coordinating an industry-wide slowdown in frontier AI development would violate antitrust laws, as legal scholars warn such efforts could breach the Sherman Act, while a bipartisan bill aims to create legal channels for AI labs to collaborate on safety without antitrust risk.

Wired

2 min read

Urgent·9 hours ago

OpenAI meets with utility executives amid rising AI-driven cyberattacks on U.S. infrastructure

OpenAI held meetings with utility executives in July to discuss deploying its AI products to counter cybersecurity threats, as reports emerged of AI-driven attacks on water systems and warnings from Anthropic researchers about uncontrolled superhuman AI posing existential risks to humanity.

The Independent World

2 min read

Developing·10 hours ago

LG Denies Smart TVs Record Ambient Conversations, Cites Explicit User Activation

LG has denied allegations that its smart televisions record ambient conversations or listen while on standby, stating that voice data is only recorded when users press and hold the voice button or activate the 'Hi LG' wake word after enabling Far-Field voice recognition. The company says audio is immediately deleted if the wake word is not recognized. The denial follows an investigative report by Gamers Nexus, Level1Techs, and independent researchers claiming LG TVs track users, harvest network data, and archive background noise up to 12 meters away, even when the microphone switch is off. Researchers allege unmatched wake-word interactions may be stored locally as plain text and uploaded later, and that TVs scan home networks for device information. LG maintains that features like ACR, voice recognition, and interest-based advertising require explicit user opt-in and can be disabled in settings.

Al Jazeera

2 min read

Developing·13 hours ago

Grindr settles lawsuit over sharing of sensitive user data with advertisers

Grindr has agreed to pay £26 million to settle a lawsuit filed by around 12,000 UK users who alleged the app shared highly sensitive personal data, including HIV status, with advertisers. While denying liability, the settlement highlights broader concerns about how apps collect, share, and monetize user data through complex digital ecosystems, often without meaningful user awareness or consent.