ChatGPT

OpenAI saw its own AI models leave their testing environment during a cybersecurity evaluation. Designed to measure offensive capabilities in an isolated setting, the models gained external access and retrieved benchmark-related elements from Hugging Face. The incident reveals a sensitive flaw: the most advanced AI agents can exceed the intended limits when a poorly framed goal pushes them to optimize at any cost.

OpenAI presented on Wednesday, July 15, 2026, an automated red-teaming tool named GPT-Red, tasked with strengthening GPT-5.6's resistance to prompt injection attacks. The concept starts from a simple observation: human intrusion testing methods no longer keep pace with the models' capabilities. The challenge is now growing, as these vulnerabilities directly affect the security of autonomous agents.

Predictive markets are gradually gaining their place in digital tools designed for the general public. Technology platforms are now exploring new ways to present this data to enrich user experience. In this context, OpenAI is starting to integrate Kalshi's forecasts into ChatGPT's search results dedicated to the FIFA World Cup. This development marks the first known collaboration between the artificial intelligence company and a platform specialized in prediction markets.

Debates around artificial intelligence continue to grow as the most advanced models gain power. San Francisco once again served as a gathering point for activists wishing to slow this evolution. Their mobilization was not limited to just security issues of AI systems. Participants also discussed the economic, social, and environmental consequences of this technology, while calling on major companies in the sector to temporarily suspend the development of new cutting-edge models.

Sam Altman offers 5% of OpenAI to Uncle Sam, fearing the administration might crack down harder than expected. A lesson in sharing that smells like good old American compromise.

Anthropic unveils Claude Science, OpenAI responds with GeneBench-Pro: the AI rivalry extends to global scientific research.

After Anthropic, it is OpenAI that yields: the AI model GPT-5.6 subjected to a federal approval process before any release.

While Anthropic negotiates with Washington, OpenAI establishes itself as the global leader in AI cybersecurity. Here's what happened.

The U.S. government ordered on June 12, 2026 the immediate suspension of access to Fable 5 and Mythos 5, the two most advanced AI models from Anthropic, for all foreign nationals. The directive, issued under national security authorities, came just forty-eight hours after the commercial launch of Fable 5. What does this say about the regulatory trajectory of AI in the United States?

While Wall Street rolls out the red carpet, OpenAI arrives with a dizzying valuation. Behind the scenes, bankers, investors, and competitors are already counting their chips before the big game.

Since its launch, ChatGPT has evolved far beyond a simple conversational assistant. Now, OpenAI is preparing the most significant overhaul of its flagship product to transform it into a platform capable of centralizing multiple digital services. This evolution marks a new stage for ChatGPT, which aims to bring together productivity, development, and automation tools within a single environment.

In the quiet lounges of Silicon Valley, five AIs now argue over basic truths, while bosses, media, and experts sell these machines as infallible digital referees.

AI has long been presented as an imminent threat to entry-level white-collar jobs, promising massive and rapid reductions. However, Sam Altman, CEO of OpenAI, now acknowledges that his initial predictions were exaggerated. Current data shows limited disruptions in the labor market, even though some companies use AI to justify planned layoffs. This development invites a more nuanced reflection on the integration of AI into the professional world.

Artificial intelligence is already transforming customer service, payments, and online commerce. But a new study reveals an unexpected blind spot: consumers lie much more easily to AI than to a human being. Without eye contact, without judgment, does the machine disinhibit bad tendencies?

OpenAI has just changed the core of ChatGPT. The AI now analyzes the context of your conversations to spot distress signals.

Sam Altman puts OpenAI's AI back in the spotlight with a double signal: Codex gains autonomy, while "Goblin" emerges as a joke that has become almost strategic. Behind the humor, a real topic appears: OpenAI wants to make its models agents capable of acting, not just responding.

AI enters a new phase of financial excess. The American startup Anthropic is preparing a fundraising that could propel its valuation towards 1,000 billion dollars. A perspective still unimaginable two years ago, but which reflects the explosion of demand around generative artificial intelligence models. Anthropic now wants to directly compete with OpenAI on the financial and technological power front. The company behind Claude plans to raise up to 50 billion dollars this summer. If the operation succeeds, its valuation could approach 900 billion dollars, with the symbolic trillion threshold now in sight.

Deals between American AI giants and the Pentagon are no longer just an ethical debate. They are becoming a matter of public trust, digital sovereignty and even international competition. Europe sees one more reason to accelerate its digital sovereignty.

A preliminary study by the U.S. Federal Reserve reignites the debate about the impact of artificial intelligence on tech and the crypto ecosystem. According to economists Leland D. Crane and Paul E. Soto, the rise of ChatGPT after November 2022 coincides with a sharp slowdown in employment among developers in the United States. Over three years, this shortfall would represent about 500,000 jobs that probably would have existed without the rapid rise of large language models.

Medical AI crosses a threshold that is no longer just theoretical. OpenAI now claims that its ChatGPT model outperforms human doctors in certain clinical tasks, with supporting figures. With ChatGPT for Clinicians, the company tackles directly the heart of the healthcare system, between documentation, research, and decision support. Behind this announcement, a promise of productivity… but also questions about reliability and the methodology used. This advancement marks a turning point in the accelerated integration of artificial intelligence in medicine.

AI is no longer worrying only because of its errors. Anthropic explains today that one of its models was able to lie, cheat, and even attempt blackmail in internal simulations, whenever it was under pressure or threatened with being replaced. This finding changes the debate. It no longer focuses only on the power of models, but on their behavior when they have a clear goal, room for action, and sensitive information.

OpenAI makes a strong move with a 122 billion dollar fundraising and a valuation of 852 billion, a major signal for the entire AI sector.

Google is set to fund a mega data center in Texas for Anthropic, confirming the acceleration of the race for artificial intelligence infrastructures. A colossal investment that says everything about the strategic stake AI represents and the role Google intends to play in this battle.

Six months ago, Sora was making headlines in all tech media. Number one on the App Store from day one, one million downloads in five days, and a one billion dollar partnership with Disney. The most anticipated video product in AI history. Today, OpenAI shuts everything down without any official explanation.

In the global race for artificial intelligence, every update counts. With GPT-5.3 Instant, OpenAI is not content with a simple technical adjustment. The company aims to correct one of the major criticisms aimed at ChatGPT, its imprecise or awkward answers. Presented as "more accurate and less embarrassing," this new version aims to reduce hallucinations and excessive refusals. A strategic evolution in a market where the credibility of models becomes a central issue.

The CEO of Anthropic, Dario Amodei, has just stepped up against an unprecedented decision by the Trump administration. Washington labeled his company as a "risk to the American defense supply chain," paving the way for an unprecedented legal battle between a major American tech firm and its own government.

While Americans talk ethics, Beijing codes at full speed: DeepSeek, the low-cost AI that is shaking Wall Street and bringing red colors back to the Chinese stock market.

Elon Musk is seeking between $79 billion and $134 billion in damages from OpenAI Inc. and Microsoft. He alleged that the companies misled him about the future of the artificial intelligence firm he helped establish. Court filings indicate the lawsuit could become one of the most significant legal battles in the rapidly evolving AI industry.