News reporting focused on AI and machine learning, covering the companies behind these technologies, their real-world applications, and the ethical concerns they raise. This includes areas like generative AI (large language models, text-to-image and video), speech tech, and predictive analytics.
A security researcher built a deliberately vulnerable book review app and spent $1,500 testing 16 different large language models on their ability to exploit a common misconfiguration. GPT-5.5 succeeded 7 out of 10 times, while several models like Gemini and Claude hit security guardrails or never found the right attack vector.
Deepseek has made its 75 percent discount on the flagship Deepseek V4 Pro model permanent, slashing output token prices to $0.87 per million versus $30 for GPT-5.5. The move intensifies the price war between Chinese and Western AI labs, with Deepseek offering output tokens at roughly 34.5 times less than OpenAI's top model. The permanent pricing, originally set to expire May 31, 2026, also undercuts Anthropic's Opus 4.7 by a similar margin on output tokens.
OpenAI is testing a feature for US Pro subscribers to connect bank accounts to ChatGPT and get analysis based on real financial data. The tool uses Plaid for connections to over 12,000 institutions and defaults to the GPT-5.5 Thinking model. It provides spending breakdowns and savings suggestions but offers read-only access with privacy controls.
OpenAI has introduced personal finance features for ChatGPT Pro subscribers in the US, enabling connections to bank accounts through Plaid. Users can analyze spending, view portfolios, and plan finances. The tools follow the acquisition of Hiro and use the improved GPT-5.5 model.
Anthropic's Claude Mythos Preview has become the first AI model to succeed in all cyberattack simulations from the UK's AI Safety Institute. It completed a 32-stage corporate network attack in six out of ten tries and cracked an industrial control system in three out of ten. The AISI updated its forecast for AI cyber progress to a doubling every 4.7 months, which models like Mythos and GPT-5.5 have already beaten.
OpenAI has updated ChatGPT to use GPT-5.5 Instant as the default model, cutting hallucinations by 52.5% on topics like medicine, law, and finance. The model shows gains across benchmarks in math, science, and visual reasoning. A new memory sources feature lets users see and edit the personal context used in responses.
OpenAI has launched GPT-5.5 Instant, replacing GPT-5.3 Instant as the default model for ChatGPT. The update cuts down hallucinations in fields like law, medicine, and finance while keeping low latency. It also boosts performance on math and multimodal benchmarks, adds context tools, and shows memory sources to users.
An analysis by the ARC Prize Foundation of 160 game runs on the ARC-AGI-3 benchmark reveals three systematic reasoning errors in OpenAI's GPT-5.5 and Anthropic's Opus 4.7. Both models score below 1 percent, with GPT-5.5 at 0.43 percent and Opus 4.7 at 0.18 percent. The errors include missing the big picture from local details, confusing new environments with training games, and failing to verify successful strategies.
OpenAI CEO Sam Altman criticized Anthropic for restricting its cybersecurity tool Mythos to select users. Now OpenAI plans a similar limited rollout for its competing tool, GPT-5.5 Cyber, starting with critical cyber defenders. The company requires applicants to submit credentials and intended uses via its website.
OpenAI has integrated its separate Codex programming model into GPT-5.5, eliminating the standalone coding line that began with GPT-5.4. Romain Huet, Head of Developer Experience at OpenAI, notes that GPT-5.3 from early February marks the final independent Codex release. The new model shows advances in agentic coding, reduced token use, and higher API costs.
OpenAI released a prompting guide for GPT-5.5, advising developers to avoid reusing prompts from older models like GPT-5.2 or GPT-5.4. The company states that detailed instructions from past versions can hinder the new model's abilities. Instead, users should craft minimal prompts focused on outcomes and follow a seven-part structure starting with role definitions.
OpenAI added GPT-5.5 and GPT-5.5 Pro to its Chat Completions and Responses APIs on April 23, 2026. These models support a 1M token context window, image inputs, and various tools. The changelog details numerous updates from 2023 to 2026, including image, video, audio models, agent tools, and API improvements.