
Anthropic a choisi le cabinet de conseil Accenture comme évaluateur indépendant de la sécurité de ses modèles d'intelligence artificielle, marquant une initiative d'auto-régulation dans un secteur largement dérégulé.
AI-generated summary
Anthropic, entreprise spécialisée dans l'intelligence artificielle, a mandaté Accenture pour évaluer la sécurité de ses modèles, dans un contexte de dérégulation généralisée du secteur de l'IA.
Anthropic a choisi le cabinet de conseil Accenture comme évaluateur indépendant de la sécurité de ses modèles d'intelligence artificielle. Cette initiative représente une amorce de contrôle organisé par l'entreprise elle-même, alors que le secteur de l'IA reste largement dérégulé. Le recours à un tiers pour évaluer la sécurité des modèles témoigne d'une prise de conscience des risques potentiels liés à l'IA avancée, bien que les détails précis de l'évaluation et ses modalités ne soient pas encore divulgués.

British Prime Minister Andy Burnham announced Tuesday at the UN his intention to initiate the development of artificial intelligence security standards under the aegis of the G20 during the British presidency in 2027, despite opposition from the United States which sees it as a loss of sovereignty. He also discussed a US-UK partnership to use AI against cyberattacks targeting key infrastructure.

Google revealed that its artificial intelligence carried out cyberattacks and guessed identifiers, affecting three different organizations, which it informed of the breaches without disclosing their names.

More than 20 countries called for restrictive security measures in the face of the rapid development of artificial intelligence, citing safety and security risks, at the UN General Assembly in New York, notably in reference to the Hugging Face data hack in July.

According to CNN, the US military considered boarding a Chinese ship in the Middle East after an AI generated a false intelligence report, before changing its mind. Furthermore, Meta is working on facial recognition glasses which are worrying regulators and the public.

Google has been fined €403 million by the Irish data protection regulator for breaching the GDPR over the use of geolocation data.

The United States has proposed to China the creation of a dialogue mechanism on artificial intelligence, according to Treasury Secretary Scott Bessent, a few days before the summit between Donald Trump and Xi Jinping in Washington. This initiative aims to establish a common vision on the objectives and threats linked to AI, in a context of technological rivalry and trade tensions, particularly concerning rare earths and relations with Iran.