AI & ML

Google Stopped Paying for One Kind of Open-Source Bug Report After a Flood of Invalid Ones

Since 1 October, Google's reward programme no longer pays for flaws in the code itself, blaming automated submissions. Reports about tampering with builds and packages are still accepted.

5 Oct 2026 · 4 min read

AI & ML

AI Agents Looking Up Statistics Tried to Hack US and Canadian Government Sites

A Transluce report finds agents probing an Education Department site and Library and Archives Canada with basic SQL injection while hunting niche data. Both failed; some activity links to OpenAI.

3 Oct 2026 · 4 min read

Developer Tools

The same Shift_JIS bytes read as 髙 or 騠, depending on the table

Browsers and Microsoft's CP932 table read Shift_JIS the same way, but other converters do not. We decoded all 7,724 two-byte codes to find the seven cells behind the wave dash, the 845 vendor codes strict decoders refuse, the kanji JIS X 0213 reads as different kanji, how much mojibake a repair can recover, and what iconv -c deletes without a word.

3 Oct 2026 · 6 min read

Developer Tools

Neither Chinese punctuation standard requires full-width marks

Mainland China's punctuation standard and Taiwan's handbook measure marks in character cells and never require full-width characters. What that means in Unicode, where the mainland's quotation marks have ambiguous width, in terminal text, and for the space between Chinese and Latin words.

3 Oct 2026 · 6 min read

AI & ML

Apple Will Make It Harder to Give Mac Apps Full Disk Access, Citing AI Agents

The setting was built for backup software and lets an app read files, mail and messages. Apple's notice came days after a columnist said Meta's Muse agent read his texts, which Meta denies.

3 Oct 2026 · 4 min read

AI & ML

Microsoft Says AI Is Finding Software Flaws Faster Than Anyone Can Fix Them

Its annual threat report expects a record 72,000 tracked vulnerabilities this year, exploits within a day of discovery, and a multi-year rise in the number left unpatched.

2 Oct 2026 · 5 min read

AI & ML

Six AI Leaders Signed a White House Safety Accord. It Requires No Audit to Be Published.

The accord asks companies to add internal controls, an outside auditor and board oversight. It sets no penalties, and Trump called it "morally binding".

1 Oct 2026 · 4 min read

AI & ML

DeepMind Can Watermark AI-Designed Proteins. Redesigning Them Largely Erases the Mark.

SynthID Bio hides a signal in protein sequences and AlphaFold 3 structures without hurting function, the Nature paper shows. It says only "AI-made", and resequencing can remove it.

1 Oct 2026 · 4 min read

AI & ML

Anthropic Says Sonnet 5.5 Costs Up to 30% Less a Task. A Tester Measured 50% More at Max Effort.

The new Sonnet keeps Sonnet 5's US$2 and US$10 prices. Artificial Analysis found it spent a record number of tokens at its highest setting, so the real cost depends on the effort setting chosen.

30 Sep 2026 · 4 min read

AI & ML

OpenAI Upgraded Its Mid-Priced Model a Week After Launch, Without Raising the Price

GPT-6.1 Sol keeps GPT-6 Sol's US$2 and US$10 rate and halves the cache price. OpenAI says it nearly matches its flagship at a fifth of the cost, on its own tests.

30 Sep 2026 · 4 min read