Linux fréttir

An anonymous reader quotes a report from Ars Technica: The system prompt for OpenAI's Codex CLI contains a perplexing and repeated warning for the most recent GPT model to "never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures unless it is absolutely and unambiguously relevant to the user's query." The explicit operational warning was made public last week as part of the latest open source code for Codex CLI that OpenAI posted on GitHub. The prohibition is repeated twice in a 3,500-plus word set of "base instructions" for the recently released GPT-5.5, alongside more anodyne reminders not to "use emojis or em dashes unless explicitly instructed" and to "never use destructive commands like 'git reset --hard' or 'git checkout --' unless the user has clearly asked for that operation." Separate system prompt instructions for earlier models contained in the same JSON file do not contain the specific prohibition against mentioning goblins and other creatures, suggesting OpenAI is fighting a new problem that has popped up in its latest model release. Anecdotal evidence on social media shows some users complaining about GPT's penchant for focusing on goblins in completely unrelated conversations in recent days. Update: OpenAI has published a blog post explaining "where the goblins came from." In short, a training signal meant to encourage its "Nerdy" personality accidentally rewarded creature-heavy metaphors, causing words like "goblins" and "gremlins" to spread beyond that personality into broader model behavior. OpenAI says it has since retired the Nerdy personality, removed the goblin-friendly reward signal, and filtered creature-word examples from training data to keep the quirk from resurfacing in inappropriate contexts.

Read more of this story at Slashdot.

Lock-in worries threaten to dampen the E7 launch party

Lock-in worries threaten to dampen the E7 launch party

The Coalition for Fair Software Licensing has published research showing that US workers reckon Microsoft is using its productivity tools to lock their employers into the company's AI services.…

Concerns over new rules might stop customers from adopting innovations – including AI – that connect to SAP systems

Concerns over new rules might stop customers from adopting innovations – including AI – that connect to SAP systems

An influential SAP user group has criticized the vendor's API policy update, saying it lacks clarity and potentially prevents users from starting new projects and innovating on their SAP platforms.…

But why did those fans go away in the first place, Satya?

But why did those fans go away in the first place, Satya?

Microsoft boss Satya Nadella told investors during an earnings call last night that the company needs to "win back" its fans.…

AI boom splits between companies hoarding eyeballs and those actually charging for them

AI boom splits between companies hoarding eyeballs and those actually charging for them

Anthropic is pulling in more LLM revenue than OpenAI, despite having a fraction of the users.…

Turns out the real problem is not AI but staff still clicking on dodgy emails from 'IT support'

Turns out the real problem is not AI but staff still clicking on dodgy emails from 'IT support'

Nearly half of UK businesses are still getting breached, and in many cases, the attacker's big breakthrough is an employee clicking "sure, why not" on a fake login page.…

Just in time for the Trump-Xi summit

Just in time for the Trump-Xi summit

Exclusive A novel China-linked threat group infiltrated more than a dozen critical networks in Poland, Asian countries, and possibly beyond, beginning in December 2024 and with activity uncovered as recently as this month.…

Longtime Slashdot reader schwit1 shares a report from ZeroHedge: The Justice Department on Tuesday sued Cloudera, accusing the enterprise data and artificial intelligence company of deliberately engineering a hiring process that excluded American workers from at least seven lucrative technology positions while the firm pursued permanent residency sponsorship for foreign workers on temporary visas. In a 14-page complaint filed with the Office of the Chief Administrative Hearing Officer, the department's Civil Rights Division alleges that Cloudera, from March 31, 2024, through at least January 28, 2025, instructed job candidates to submit applications to a dedicated email address, amerijobpostings@cloudera.com, that rejected all external messages with an automated bounce-back error. The company did not advertise the roles on its public careers website or accept applications through its standard portal, as it did for non-sponsorship positions. Cloudera then attested to the Department of Labor that it could not locate any qualified U.S. workers for the roles, which paid between approximately $180,000 and $294,000 annually, according to the filing. The positions included a Product Manager role in Santa Clara, California, with a listed salary range of $170,186 to $190,000. The case marks one of the most detailed enforcement actions under the Justice Department's Protecting U.S. Workers Initiative, which was relaunched last year and has already produced 10 settlements targeting employers accused of discriminating against American workers in favor of temporary visa holders. "Employers cannot use the PERM sponsorship process as a backdoor for discriminating against U.S. workers," Assistant Attorney General Harmeet K. Dhillon of the Civil Rights Division said in a statement. "The Division will not hesitate to sue companies who intentionally deter U.S. workers from applying to American jobs."

Read more of this story at Slashdot.

Emergency patches out now for those managing the millions of domains assumed to be affected

Emergency patches out now for those managing the millions of domains assumed to be affected

Emergency patches are available for a critical vulnerability in cPanel and WHM that allows attackers to bypass authentication and gain root access to servers managed using it.…

Federation warns members to ditch work devices off duty as force uses AI to probe 600+ cops

Federation warns members to ditch work devices off duty as force uses AI to probe 600+ cops

London cops are being told by their staff association to be "extremely cautious" about carrying work devices off duty, after the Metropolitan Police Service (MPS) deployed Palantir's technology to investigate hundreds of its own officers.…

Investigation finds no single cause for soldiers falling ill, just bad bolts, cold air, and apparently the soldiers themselves

Investigation finds no single cause for soldiers falling ill, just bad bolts, cold air, and apparently the soldiers themselves

Britain's notorious Ajax armored vehicles are being accepted back from the manufacturer after investigations found no single cause for the symptoms plaguing crews, meaning soldiers will need to grin and bear it.…