OpenAI Blog
AI news source. Articles are auto-selected and adapted by Hamidun News editors.
Latest publications
How AI agents defend against prompt injections
ChatGPT has learned to resist manipulation: how AI agents block dangerous commands and protect user data.
Rakuten doubled bug-fixing speed with AI
Japanese tech giant Rakuten deployed an AI coding agent and cut system recovery time by 50%.
Descript and OpenAI: how to scale multilingual video dubbing
Descript uses OpenAI models for automatic video dubbing into dozens of languages. The system optimizes translation not only for meaning but also for timing, so dubbed speech sounds natural.
Reasoning Models Can't Hide Their Thoughts — And That's Good
OpenAI introduced the CoT-Control tool and discovered: reasoning models are nearly unable to control their chains of thought. Paradoxically, this very 'weakness' could become a key mechanism for AI safety.
Axios: how AI is helping local journalism survive
Axios's operations director explained how the newsroom uses AI to support local reporters and scale high-quality journalism.
AI helped compute graviton amplitudes in quantum gravity
A new preprint extends the single-minus amplitude method to gravitons — and GPT-5.2 Pro played a key role in deriving and verifying the results.
OpenAI introduces new tools to protect users' mental health
The company behind ChatGPT unveiled parental controls and improved algorithms for detecting signs of stress to increase the safety of AI systems.
OpenAI and Microsoft update partnership: what is behind the joint statement
OpenAI and Microsoft issued a joint statement on continuing their cooperation in research, engineering, and product development. Behind the concise wording lies a complex history in the relationship between the two giant
OpenAI and Amazon join forces: agents gain persistent memory in Bedrock
Amazon and OpenAI have introduced Stateful Runtime Environment, a runtime environment for AI agents with persistent memory, orchestration, and secure execution of multi-step tasks. This changes the game for enterprise AI
OpenAI and U.S. national lab aim to speed up bureaucracy with AI
Pacific Northwest National Laboratory and OpenAI have created DraftNEPABench, a benchmark that assesses AI agents' ability to speed up federal environmental reviews — and early results suggest timelines could be reduced
OpenAI reveals new AI abuse schemes: threats are becoming more complex
OpenAI has published its February threat report: malicious actors are increasingly combining language models with web platforms and social media, creating multi-layered attacks. The company explained what this means for
OpenAI drops the leading benchmark for evaluating code — and that changes the rules of the game
OpenAI announced that it will no longer use SWE-bench Verified — one of the most cited benchmarks for evaluating AI coding models. The reason: contaminated data, leakage into training datasets, and flawed tests. The comp
OpenAI launches Frontier Alliance Partners to scale AI agents in business
OpenAI has introduced the Frontier Alliance Partners program, a partner network that will help enterprises move from pilot AI projects to full production systems with safe, scalable agents.
OpenAI and Paradigm developed a test for AI auditing of smart contracts
The new EVMbench benchmark measures how well AI agents can find and fix critical vulnerabilities in blockchain code.
OpenAI invested $7.5 million in independent AI safety research
OpenAI is funding an independent project on aligning AI with human values as the stakes rise with AGI drawing closer.
OpenAI officially enters the Indian market
Sam Altman’s company is expanding its global presence and launching initiatives to support the largest developer community.
OpenAI introduces ChatGPT Go: global access to GPT-5.2 Instant
OpenAI is launching ChatGPT Go, a new version of the service based on GPT-5.2 Instant with expanded memory and higher limits, making advanced AI more accessible worldwide.
OpenAI introduces ads: ChatGPT to become more accessible through advertising
OpenAI announced plans to test ads in ChatGPT. This will help preserve free access to AI and expand its capabilities for users around the world.
OpenAI's business model: scaling through the value of intelligence
OpenAI is reshaping its strategy by betting on deeper ChatGPT integration and new revenue sources—from advertising to compute sales.
AI as a tool for expanding human capabilities
We examine the concept of expanding human potential with AI and how technology helps bridge skills gaps and drive productivity growth across all areas of life.
Praktika uses GPT-5 to create personalized AI tutors
Startup Praktika uses the latest language models to teach foreign languages through live conversation and deep lesson personalization for each user.
OpenAI begins testing ads in ChatGPT to support the free version
OpenAI is introducing ad units into the chatbot interface. Developers say they will preserve data privacy and the objectivity of the AI model's responses.
The Age of Agents: How Codex Is Changing the Architecture of Software Development
The shift to autonomous agents requires new architectural approaches. We examine how Codex is becoming the core of systems capable of writing and testing code independently.
OpenAI introduced GPT-5.3-Codex-Spark: instant code generation and 128k context
The new GPT-5.3-Codex-Spark model promises to change how developers work. Code generation is now 15 times faster, and the context window has expanded to 128,000 tokens. Available in ChatGPT Pro.
Scaling Access: How OpenAI Manages Load for Sora and Codex
OpenAI has revealed details of the architecture of its access management system. Learn how the company combines limits and credits to keep its most resource-intensive models running stably in real time.
GPT-5.2 made a breakthrough in theoretical physics, deriving a new formula
OpenAI's AI managed to solve a complex problem in quantum chromodynamics. Scientists officially confirmed the accuracy of the neural network's calculations.
OpenAI unveiled GABRIEL: an AI tool for large-scale sociological research
OpenAI's new open-source project turns qualitative data into quantitative data, allowing researchers to analyze vast datasets with GPT.
GPT-5 in the Office: Who's Actually Working for You
Admit it, you're doing it too. You open a tab with an AI, have it rewrite that awkward email to a client or explain what your colleague meant in their endless report.
Codex CLI and agent loop: how autonomy magic actually works
It seems we've begun to forget that behind every "smart" action of a neural network stands not magic, but rather a fairly cumbersome engineering framework.
Indeed and AI: jobs now seek you (not the other way around)
You've ever wondered why job hunting in 2024 still feels like trying to find a needle in a burning haystack? Indeed's Chief Revenue Officer Maggie Halls believes this era is finally coming to an end.