Morning edition — Monday 19 October 2026

The vLLM project published v0.32.0 on October 18 at 08:50 CEST. The release notes credit 814 commits from 363 contributors.

On security, it fixes three advisories: GHSA-4hhp-h66f-j5j7 (chat template resource exhaustion), GHSA-q97c-p368-75c4 (tokenizer max_token_id) and GHSA-frvg-qw88-hww9 (P2P ZMQ socket exhaustion). CORS no longer allows every origin by default, untrusted images are limited to raster formats, shared-memory multimodal cache handles are signed and JSON schemas nested deeper than 64 levels are rejected.

Breaking changes include dropping Python 3.10, new transformers and tokenizers version pins, removal of cascade attention and of the torch.compile sequence-parallelism and AsyncTP passes, no longer reserving max_tokens from the context window, and renamed Responses API reasoning stream events. Other additions include a new POST /v1/systemone endpoint for structured decisions and NVFP4 KV-cache support on more GPU generations.

Business Insider reported at 02:07 CEST on October 19 that Meta had given a fuller explanation for locking the No Kings movement's Instagram account during its Saturday day of action. The group said the account was locked and administrators were denied access for several hours while it held rallies to encourage early voting in the US midterms.

In an X post quoted by Business Insider, Meta spokesperson Andy Stone said engineers found the spike that triggered the incorrect enforcement was mainly in the account's 'reading' behavior: network security systems detected 'the use of a Claude tool reading profiles, posts, and hashtags,' at a rate that rose about fourfold in intense bursts on October 17. Stone said the pattern resembled bot traffic, that an automated re-review restored the account an hour later, that fixing administrator access took longer, and he repeated Meta's apology.

Meta had first described the shutdown as an automatic trigger acting in error. The report does not say who was running the Claude tool or how it was configured.

According to SiliconANGLE, which cites a Sunday report in The Telegraph, Microsoft has ended an agreement to rent AI infrastructure at Nscale's planned facility in Loughton, Essex. The project had faced repeated construction delays, and UK grid officials told Nscale they could not supply adequate power to the site until the 2030s. Nscale had planned to install 23,040 Nvidia GPUs there by the first quarter of 2027, for an initial 50 megawatts of capacity.

An Nscale spokesperson told The Telegraph that its relationship with Microsoft remains 'strong and long-term', that the grid issue does not affect its other Microsoft contracts, and that it is exploring alternative power sources for Loughton. SiliconANGLE reports that Microsoft and Anthropic account for about 85% of Nscale's contract value, according to its IPO filing.

Overnight, PYMNTS relayed a Bloomberg News report that some Nscale backers are balking at an IPO valuation goal of around $30 billion and suggesting $16 billion to $19 billion instead. SiliconANGLE separately cites a target of about $35 billion for a $3 billion raise on the New York Stock Exchange under the ticker NSCL. The two figures come from different press reports and have not been confirmed by the company.

Anthropic said on October 18 (18:49 CEST) that, subject to stockholder approval, it is committing 5% of its shares to the Anthropic Public Benefit Fund, with a further 2% to be added at the board's discretion for every 50% increase in the company's value above a set baseline. The baseline has not been published.

The fund is a wholly owned LLC rather than a 501(c)(3), which Anthropic says leaves room for large donations to nonprofits or direct distributions to specific populations. It names AI's effect on the labor market as one example, building on a previously announced $200 million commitment, and says none of the fund's equity will be used for lobbying or political giving, in the US or abroad.

Combined with the co-founders' pledge to donate 80% of their equity and an employee donation-match program in which about half of staff take part, Anthropic says roughly 20% of its shares would be dedicated to the public good. It promises details on governance and the valuation milestones in the coming months. The BBC notes the move comes as Anthropic prepares a possible stock market listing, and reports that Elon Musk replied: 'I recommend giving the shares directly to the public.'

TechCrunch reported on October 18 (18:30 CEST) that a legal filing spotted by Bloomberg shows Disney agreed to make job offers to at least 60 Character.AI employees, the majority of the chatbot startup's workforce, with estimated annual compensation of $60 million. The filing says the 'vast majority' of those offered jobs accepted.

Character.AI filed for bankruptcy earlier in the week. The startup had previously said Disney agreed to buy some of its assets for $20 million, and its former CEO, Karandeep Anand, recently became Disney's first CTO. Disney's offer is a stalking-horse bid, so another buyer could still come in higher.

According to TechCrunch, the filing says that despite a 'large and engaged user base', Character.AI was hobbled by high costs and 'potential litigation-related liabilities and regulatory engagement'.