DeloIndependent🔒Center17 hr. agoOpenAI confirmed it will publicly release three new AI models—GPT-5.6 Sol, Terra, and Luna—on Thursday. The most powerful version, GPT-5.6 Sol, was originally scheduled for June but was delayed at the request of the U.S. government to assess national security risks. This announcement comes just a week after the U.S. lifted restrictions on Anthropic’s latest models, Fable and Mythos, which had been temporarily blocked due to concerns over potential security threats. According to Axios, the U.S. Commerce Department approved the wider public release of GPT-5.6 under a new oversight system for advanced AI models. Previously, access to the model was limited to a small group of verified partners whose identities were shared with U.S. authorities. The U.S. administration has tightened control over AI model releases in recent months due to fears that countries like China or Russia might exploit them for military or intelligence purposes. Anthropic noted that creating an AI model completely resistant to 'jailbreak' techniques—methods used to bypass safety restrictions—is likely impossible.
Bias read (Center): The article provides a balanced overview of the situation, mentioning both the actions of OpenAI and the U.S. government, as well as comments from Anthropic. It does not exhibit strong ideological framing or biased language.
Slo-TechIndependentCenteryesterdayAnthropic, ponudnik AI-modela Claude, je odstranil steganografske podpise iz svojega modela Claude Code, ki so skrivaj sledili uporabi modela stranim uporabnikom. Uporabnik Thereallo je odkril te podpise v kodah, ki so vključevali informacije o času, posredniškem strežniku in povezavi s Kitajsko. Anthropicov inženir Thariq Shihipar je potrdil, da so eksperimenta začeli marca, da bi preprečili nezakonito preprodajo ali destilacijo modela. Preprodaja dostopa se je dejansko dogajala. Anthropic ni edino podjetje, ki očituje Kitajce za destilacijo modelov, vendar to ne pomeni, da njihova ukrepanja niso smislena. Članek navaja, da bi podjetje lahko uporabnike obvesčilo o metodah za preprečevanje nezakonite rabe.
Bias read (Center): Člank je neutralen v svojem okviru, kljub temu, da obravnava politično relevantno temo (destilacija AI-modelov in mednarodni odnos). Ključne informacije so predstavljene brez evidentnega stranskoglasnega okvira. Anthropicov odgovor in odkritje uporabnika sta predstavljeni kot fakta, brez izražanja v
Slo-TechIndependentCenter3 days ago The U.S. wants five percent of OpenAIThe article discusses U.S. administration and OpenAI CEO Sam Altman aligning on making the United States a partial owner of OpenAI, aiming for a 5% stake. The goal is to allow the U.S. to benefit from AI advancements while avoiding regulatory restrictions on product sales. Altman suggests creating a sovereign fund similar to Norway’s or Alaska’s model, where major AI developers could donate shares. Critics argue this resembles corporate welfare and question the motives behind such proposals.
Bias read (Center): While the topic involves government ownership of a private company, the article presents both sides of the debate—supporting the idea as a way to share AI benefits and criticizing it as a form of corporate subsidy. There is no clear ideological leaning in the framing, though some commenters express左
Slovenia
