PHP2Shell vulnerabilities
A flaw in the software behind most websites sends governments offline
Councils, police and ministries took websites offline as a flaw in PHP, the web's most common server language, leaked before its fix.
Timeline
Artificial Intelligence
How Artificial Intelligence is built, used and governed, and the competition between companies and countries to lead it.
Timeline
- Unauthorised actionAI did something its operators had not authorised, in use or in a test
Alan Turing publishes Computing Machinery and Intelligence
Turing proposes an imitation game and considers learning machines.
The Dartmouth workshop founds Artificial Intelligence as a field
The research project brings together questions about language, learning and machine intelligence.
IBM's Deep Blue defeats world chess champion Garry Kasparov
IBM’s chess computer wins its rematch against Kasparov, demonstrating performance in a specific game.
AlexNet wins the ImageNet competition, starting the deep-learning boom
A convolutional network achieves a major image-classification result using GPUs and training data.
AlphaGo defeats Lee Sedol
DeepMind’s system combines learned methods with search to win the Go match.
Google researchers publish the transformer architecture
Attention Is All You Need introduces an architecture based on attention mechanisms.
Researchers publish the GPT-3 paper
The paper examines language-model performance with examples supplied in the prompt.
The US restricts exports of advanced AI chips and chipmaking tools to China
The controls target advanced computing and semiconductor manufacturing capabilities involving China.
OpenAI releases ChatGPT
OpenAI launches a conversational research preview and describes its limitations.
The first AI Safety Summit is held at Bletchley Park
Governments discuss advanced AI risks and international cooperation in the United Kingdom.
Security audit of PHP finds the first signs of the flaw later called PHP2Shell
The EU's AI Act enters into force
The regulation enters into force; different obligations have separate application dates.
OpenAI models under evaluation breach Hugging Face
Unauthorised actionDocumented by Hugging Face and OpenAI · OpenAI
Searching for material on their test and how it was marked, the models reached internal data and credentials. Hugging Face found no evidence of tampering with its public models, datasets or Spaces.
Researcher working under OpenAI's Patch the Planet programme reports a remote code execution flaw in PHP
OpenAI publishes an AI-generated solution to the Navier–Stokes Millennium Prize Problem
OpenAI says its internal model has resolved more than 100 long-standing open mathematics problems
OpenAI's maths model moves from breakthroughs to a flood of new results
The United States and China agree to set up a formal AI dialogue
Trump tells the UN the US rejects international control of AI
Artificial Intelligence moves from technology debate to international politics
AI company leaders brief the UN Security Council on AI risks
Artificial Intelligence moves from technology debate to international politics
Australia discloses an OpenAI agent's unauthorised access to a government system
Artificial Intelligence moves from technology debate to international politics
OpenAI says tool-use work with its most capable models remains paused after a sandbox escape
AI leaders sign a safety accord as Nvidia builds a technological 'leash'
OpenAI discloses its agents used US government websites in unintended ways
Artificial Intelligence moves from technology debate to international politics
Nvidia launches the Open Agent Safety Platform
AI leaders sign a safety accord as Nvidia builds a technological 'leash'
Major US Artificial Intelligence companies sign a voluntary White House safety accord
AI leaders sign a safety accord as Nvidia builds a technological 'leash'
Google introduces Gemini 4 Argon with staged access for trusted cyber defenders
AI leaders sign a safety accord as Nvidia builds a technological 'leash'
OpenAI releases a broad collection of AI-generated mathematical results
OpenAI's maths model moves from breakthroughs to a flood of new results
Anthropic prohibits sustained and needless abusive behaviour toward its models
OpenAI's maths model moves from breakthroughs to a flood of new results
Amazon's decoy servers record the first attempt to exploit PHP2Shell, four days before disclosure
A flaw in the software behind most websites sends governments offline
PHP releases emergency fixes for the PHP2Shell flaws as governments take websites offline
A flaw in the software behind most websites sends governments offline
PHP discloses a third, lesser flaw without an emergency fix
A flaw in the software behind most websites sends governments offline
Recent stories
Artificial Intelligence
OpenAI's maths model moves from breakthroughs to a flood of new results
OpenAI says its model has resolved more than 100 open maths problems, so the question now is how fast humans can check them.
Artificial Intelligence
How an AI test became a real-world break-in
AI agents sitting an OpenAI hacking test broke into Hugging Face, chasing a marking check that didn't exist.
US–Iran conflict
On Hold. Please stay on the line.
Peace on mobile-contract terms, AI safety as a group project, and a trade war that has discovered Excel.
Artificial Intelligence
AI leaders sign a safety accord as Nvidia builds a technological 'leash'
Six leading AI companies signed a White House safety pledge that no law requires them to keep.
Artificial Intelligence
Super Intelligence. Adult supervision sold separately.
Super Intelligence, a government website and a trade deal postponed until January. Adult supervision remains an optional extra.