
Researchers are warning of the unprecedented risks posed by advanced AI models from OpenAI and Anthropic, capable of identifying and exploiting security vulnerabilities at unprecedented speed, pushing companies and governments to urgently strengthen their digital defenses.
AI-generated summary
AI models like OpenAI's Astra are touted as extremely capable but raise concerns about their potential to be used to identify and exploit IT vulnerabilities at unprecedented speed.
Researchers are warning of the unprecedented risks posed by advanced AI models from OpenAI and Anthropic, capable of identifying and exploiting security vulnerabilities at unprecedented speed, pushing companies and governments to urgently strengthen their digital defenses.
“We are not ready for what comes next”: the powerful models of OpenAI and Anthropic, an unprecedented threat to global cybersecurity
DECRYPTION - Faced with AI models capable of identifying and exploiting vulnerabilities at unprecedented speed, companies and States must urgently review their defenses.
Panic of researchers, opaque architecture, disproportionate capacities: what you need to know about Astra, the new AI which takes us “into the era of superintelligence”
DECRYPTION - Capable of using your computer alone, the new OpenAI model is presented as being “the most aligned” to date. And this, despite the concern of some experts, who call Astra “the worst development ever observed in terms of AI safety”.
AI outlook — possibilities, not facts
Governments to accelerate regulations on high-risk AI models
Likely · Within months
Investments in defensive cybersecurity will increase significantly
Very likely · Within months

Summary of the threats linked to AI for cybersecurity, the conviction of the Loire private hospital after a data leak of 500,000 patients, and an alert from the Somme gendarmerie concerning “SIM swapping”.

Portrait of Daniela Amodei, president and co-founder of Anthropic, valued at $965 billion, considered one of the most powerful figures in AI.
James Dyson unveils CameraJet, an electric toothbrush equipped with a camera and AI for precise dental cleaning. The inventor discusses technical challenges, data protection and his vision of innovation in the face of counterfeiting.

German start-up Isar Aerospace successfully launched its Spectrum rocket from the Andøya base in Norway, becoming the first private European company to place a satellite in orbit, thus aiming to strengthen the continent's strategic autonomy.

The Tetris Company said it had not authorized the use of its trademark for the game "Build the Wall", visually inspired by Tetris and published by the White House on its site Thursday, inviting people to build a wall on the US-Mexico border against "zombies". Another game featuring Tom Homan was also released, drawing criticism from the NGO Frontera Federation.
Hackers have released more than a million pieces of data stolen from the Berlin city administration after an ultimatum for a two million euro ransom expired, according to RBB and Der Spiegel. Mayor Kai Wegner refused to pay the ransom announced at the end of August. The cyberattack of August 14 disrupted several municipal services for a week.