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.
Anthropic has released Claude Fable 5, a restricted version of its powerful Mythos AI system. The new model costs twice as much as the previous flagship and includes guardrails to prevent misuse in cybersecurity and biology, after the company warned Mythos could aid hackers.
Anthropic has released Claude Fable 5, its most powerful widely available AI model and the first from the previously withheld Mythos class. The company claims new safeguards make the release safe by blocking responses in high-risk areas like cybersecurity and biology, falling back to the Claude Opus 4.8 model. Pricing is double that of Opus 4.8, at $10 per million input tokens and $50 per million output tokens.
Valentina Palmiotti, known as Chompie, won top prizes at the Pwn2Own hacking competition in Berlin but says AI models like Anthropic's Claude Mythos are so powerful that even champion ethical hackers will soon struggle to compete. She predicts the window for human bug bounty hunters is closing, though some peers remain optimistic about human creativity.
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.
Anthropic is discussing a new funding round of $30 billion to $50 billion that could value the AI startup at $950 billion, up from $380 billion three months ago. The company recently launched the Mythos AI model and faces tensions with the Pentagon while reporting a $30 billion revenue run rate. Deals with Google and Amazon provide additional support amid fierce competition.
The Trump administration discussed an executive order for government review of new AI models before public release. Officials briefed representatives from Anthropic, Google, and OpenAI last week. The move follows concerns over Anthropic's powerful Mythos model, which the company withheld due to cybersecurity risks and the NSA now uses.
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.
Google intends to invest up to $40 billion in Anthropic, including $10 billion immediately at a $350 billion valuation and $30 billion more upon meeting targets. This supports Anthropic's compute needs after releasing its powerful Mythos model. The deal builds on existing partnerships amid the AI industry's compute competition.
OpenAI launched its Trusted Access for Cyber program, granting Microsoft access to its most advanced models for cybersecurity tasks. Microsoft, in exchange, dedicates its full cybersecurity team to safeguard OpenAI's models, infrastructure, and mutual customers. This move aligns with Microsoft's Secure Future Initiative and comes amid debates on AI capabilities in cyber defense, including Anthropic's Mythos model claims.
Anthropic revealed its powerful AI model Mythos, which it deemed too risky for broad release, naming 11 U.S. organizations as partners. The model quickly prompted reactions worldwide from central banks and intelligence agencies after it demonstrated abilities to detect and exploit vulnerabilities in critical software for banks, power grids, and governments. Only Britain received access outside the U.S., fueling geopolitical tensions as leaders assess risks and rivals like China and Russia view it as a major threat.
A group of unauthorized users accessed Anthropic's powerful Claude Mythos AI model using a contractor's credentials and leaked data. The model, limited to partners like Apple, Amazon, and Cisco under Project Glasswing, was used only for benign tasks. Anthropic is investigating with no signs of broader compromise.
A private Discord group has gained unauthorized access to Anthropic's Claude Mythos Preview, a powerful AI model designed for cybersecurity tasks. The access occurred on April 7, coinciding with the model's limited release announcement, and has continued for two weeks. Anthropic is investigating the incident, which involved a third-party contractor and data from a prior breach.