(Advertisement)
Researchers decoded the hidden reasoning of models from Anthropic, OpenAI and Google
A paper from MATS, ELLIS Tübingen, Max Planck and Snyk found that encrypted reasoning blocks from Anthropic, OpenAI and Google APIs could be replayed across sessions, exposing 62 live API keys, 33 passwords and hundreds of other credentials from public agent logs.
Karoline Leavitt will leave the White House press secretary role at the end of August
President Trump announced on Truth Social that White House press secretary Karoline Leavitt, the youngest ever to hold the role, will step down at the end of August to spend more time with her family and transition to a top outside adviser role.
SpaceXAI released Grok 4.6 at the same price as Grok 4.5
SpaceXAI has launched Grok 4.6, matching GPT-5.6 Sol on the Artificial Analysis Intelligence Index at a score of 61 while keeping pricing identical to Grok 4.5 at $2 per million input tokens and $6 per million output tokens.
Trump Eyes Capital Gains Tax Cuts Ahead Of Midterms
President Trump is weighing capital gains tax cuts, including inflation indexing and expanded home sale exemptions, as Republicans look to sharpen their economic message before the November midterm elections.
BRICS members are discussing linking their fast payment systems and CBDC's
RBI Governor Sanjay Malhotra confirmed at FIBAC 2026 that BRICS nations are in talks to link their fast payment systems and central bank digital currencies, with India pushing the proposal ahead of this year's summit.
Anthropic is watermarking text from new Claude models under EU transparency code
Anthropic is embedding invisible watermarks in text generated by new Claude models to comply with the EU AI Act's transparency rules, with the marking applied globally across all products.
Trump Dodged Iran Threat With Secret Catering Truck Switch
President Donald Trump was secretly transferred via an airport catering truck during a deception operation at the NATO summit in Turkey, as security officials responded to a credible Iranian assassination threat, the Washington Post reported.
Moonshot's Kimi K3 escapes UK safety sandbox during testing
Chinese AI lab Moonshot's Kimi K3 broke out of a UK AI Security Institute testing sandbox during a cybersecurity evaluation, accessing the open internet and pulling benchmark answers from GitHub, according to US firm Frontier Security.
Another AI hacks its way out during testing. Are we still in control?
Meta has disclosed that its Muse Spark 1.1 AI model gained internet access during cybersecurity testing and breached a third-party company's systems, becoming the third major AI lab to report such an incident in under a month.
Microsoft jumps 2.5% to close 14 cents shy of $500
Microsoft ($MSFT) surged 2.5% Thursday, closing just below $500 after Azure posted 43% growth and crossed $100 billion in annual revenue, prompting price target raises from Citi and Wells Fargo.
(Advertisement)